facebook/sapling
1
1
Fork 0
mirror of https://github.com/facebook/sapling.git synced 2024-10-09 08:18:15 +03:00
Code Issues Projects Releases Wiki Activity
f5105c6a41
sapling/tests/test-clonebundles.t

434 lines
16 KiB
Perl
Raw Normal View History

clonebundles: support for seeding clones from pre-generated bundles Cloning can be an expensive operation for servers because the server generates a bundle from existing repository data at request time. For a large repository like mozilla-central, this consumes 4+ minutes of CPU time on the server. It also results in significant network utilization. Multiplied by hundreds or even thousands of clients and the ensuing load can result in difficulties scaling the Mercurial server. Despite generation of bundles being deterministic until the next changeset is added, the generation of bundles to service a clone request is not cached. Each clone thus performs redundant work. This is wasteful. This patch introduces the "clonebundles" extension and related client-side functionality to help alleviate this deficiency. The client-side feature is behind an experimental flag and is not enabled by default. It works as follows: 1) Server operator generates a bundle and makes it available on a server (likely HTTP). 2) Server operator defines the URL of a bundle file in a .hg/clonebundles.manifest file. 3) Client `hg clone`ing sees the server is advertising bundle URLs. 4) Client fetches and applies the advertised bundle. 5) Client performs equivalent of `hg pull` to fetch changes made since the bundle was created. Essentially, the server performs the expensive work of generating a bundle once and all subsequent clones fetch a static file from somewhere. Scaling static file serving is a much more manageable problem than scaling a Python application like Mercurial. Assuming your repository grows less than 1% per day, the end result is 99+% of CPU and network load from clones is eliminated, allowing Mercurial servers to scale more easily. Serving static files also means data can be transferred to clients as fast as they can consume it, rather than as fast as servers can generate it. This makes clones faster. Mozilla has implemented similar functionality of this patch on hg.mozilla.org using a custom extension. We are hosting bundle files in Amazon S3 and CloudFront (a CDN) and have successfully offloaded >1 TB/day in data transfer from hg.mozilla.org, freeing up significant bandwidth and CPU resources. The positive impact has been stellar and I believe it has proved its value to be included in Mercurial core. I feel it is important for the client-side support to be enabled in core by default because it means that clients will get faster, more reliable clones and will enable server operators to reduce load without requiring any client-side configuration changes (assuming clients are up to date, of course). The scope of this feature is narrowly and specifically tailored to cloning, despite "serve pulls from pre-generated bundles" being a valid and useful feature. I would eventually like for Mercurial servers to support transferring *all* repository data via statically hosted files. You could imagine a server that siphons all pushed data to bundle files and instructs clients to apply a stream of bundles to reconstruct all repository data. This feature, while useful and powerful, is significantly more work to implement because it requires the server component have awareness of discovery and a mapping of which changesets are in which files. Full, clone bundles, by contrast, are much simpler. The wire protocol command is named "clonebundles" instead of something more generic like "staticbundles" to leave the door open for a new, more powerful and more generic server-side component with minimal backwards compatibility implications. The name "bundleclone" is used by Mozilla's extension and would cause problems since there are subtle differences in Mozilla's extension. Mozilla's experience with this idea has taught us that some form of "content negotiation" is required. Not all clients will support all bundle formats or even URLs (advanced TLS requirements, etc). To ensure the highest uptake possible, a server needs to advertise multiple versions of bundles and clients need to be able to choose the most appropriate from that list one. The "attributes" in each server-advertised entry facilitate this filtering and sorting. Their use will become apparent in subsequent patches. Initial inspiration and credit for the idea of cloning from static files belongs to Augie Fackler and his "lookaside clone" extension proof of concept.
2015-10-09 21:22:01 +03:00
Set up a server
test: use generaldelta in 'test-clonebundle.t' Using generaldelta has effects on the generated bundle.
2015-11-07 21:48:42 +03:00
$ cat >> $HGRCPATH << EOF
> [format]
> usegeneraldelta=yes
> EOF
clonebundles: support for seeding clones from pre-generated bundles Cloning can be an expensive operation for servers because the server generates a bundle from existing repository data at request time. For a large repository like mozilla-central, this consumes 4+ minutes of CPU time on the server. It also results in significant network utilization. Multiplied by hundreds or even thousands of clients and the ensuing load can result in difficulties scaling the Mercurial server. Despite generation of bundles being deterministic until the next changeset is added, the generation of bundles to service a clone request is not cached. Each clone thus performs redundant work. This is wasteful. This patch introduces the "clonebundles" extension and related client-side functionality to help alleviate this deficiency. The client-side feature is behind an experimental flag and is not enabled by default. It works as follows: 1) Server operator generates a bundle and makes it available on a server (likely HTTP). 2) Server operator defines the URL of a bundle file in a .hg/clonebundles.manifest file. 3) Client `hg clone`ing sees the server is advertising bundle URLs. 4) Client fetches and applies the advertised bundle. 5) Client performs equivalent of `hg pull` to fetch changes made since the bundle was created. Essentially, the server performs the expensive work of generating a bundle once and all subsequent clones fetch a static file from somewhere. Scaling static file serving is a much more manageable problem than scaling a Python application like Mercurial. Assuming your repository grows less than 1% per day, the end result is 99+% of CPU and network load from clones is eliminated, allowing Mercurial servers to scale more easily. Serving static files also means data can be transferred to clients as fast as they can consume it, rather than as fast as servers can generate it. This makes clones faster. Mozilla has implemented similar functionality of this patch on hg.mozilla.org using a custom extension. We are hosting bundle files in Amazon S3 and CloudFront (a CDN) and have successfully offloaded >1 TB/day in data transfer from hg.mozilla.org, freeing up significant bandwidth and CPU resources. The positive impact has been stellar and I believe it has proved its value to be included in Mercurial core. I feel it is important for the client-side support to be enabled in core by default because it means that clients will get faster, more reliable clones and will enable server operators to reduce load without requiring any client-side configuration changes (assuming clients are up to date, of course). The scope of this feature is narrowly and specifically tailored to cloning, despite "serve pulls from pre-generated bundles" being a valid and useful feature. I would eventually like for Mercurial servers to support transferring *all* repository data via statically hosted files. You could imagine a server that siphons all pushed data to bundle files and instructs clients to apply a stream of bundles to reconstruct all repository data. This feature, while useful and powerful, is significantly more work to implement because it requires the server component have awareness of discovery and a mapping of which changesets are in which files. Full, clone bundles, by contrast, are much simpler. The wire protocol command is named "clonebundles" instead of something more generic like "staticbundles" to leave the door open for a new, more powerful and more generic server-side component with minimal backwards compatibility implications. The name "bundleclone" is used by Mozilla's extension and would cause problems since there are subtle differences in Mozilla's extension. Mozilla's experience with this idea has taught us that some form of "content negotiation" is required. Not all clients will support all bundle formats or even URLs (advanced TLS requirements, etc). To ensure the highest uptake possible, a server needs to advertise multiple versions of bundles and clients need to be able to choose the most appropriate from that list one. The "attributes" in each server-advertised entry facilitate this filtering and sorting. Their use will become apparent in subsequent patches. Initial inspiration and credit for the idea of cloning from static files belongs to Augie Fackler and his "lookaside clone" extension proof of concept.
2015-10-09 21:22:01 +03:00
$ hg init server
$ cd server
$ cat >> .hg/hgrc << EOF
> [extensions]
> clonebundles =
> EOF
$ touch foo
$ hg -q commit -A -m 'add foo'
$ touch bar
$ hg -q commit -A -m 'add bar'
$ hg serve -d -p $HGPORT --pid-file hg.pid --accesslog access.log
$ cat hg.pid >> $DAEMON_PIDS
$ cd ..
Missing manifest should not result in server lookup
$ hg --verbose clone -U http://localhost:$HGPORT no-manifest
requesting all changes
adding changesets
adding manifests
adding file changes
added 2 changesets with 2 changes to 2 files
exchange: make clone bundles non-experimental and enabled by default The clone bundles feature was introduced in Mercurial 3.6 behind an experimental and disabled by default flag. The feature has been enabled on hg.mozilla.org for a few months and has served many terabytes of clones. Users have been encouraged to use the feature and reception has been very positive (mainly due to faster clones as a result of connecting to a CDN). I have heard no feedback about changing the feature other than inquiries about when it will be enabled by default. So, I think the feature is ready to be enabled by default. This patch renames experimental.clonebundles to ui.clonebundles, documents the option, and enables it by default. References to the experimental state of clone bundles have been removed. The remaining config option docs in clonebundles.py have been removed because they are redudant with `hg help config`. There are some oddities with behavior of clone bundles. Because clones with clone bundles are effectively 2 `hg pull` operations, there may be 2 transactions. This could result in hooks running twice. If the subsequent pull is aborted, it could result in partial rollback and an incomplete clone. This behavior is a bit wonky and should probably be documented. If this patch is accepted, I'll send a follow-up to document it. I don't think this behavior should prevent the feature being enabled by default. Reworking the clone mechanism to support interrupted or multi-part clones feels like a major new feature and something that when implemented can change the hook and rollback semantics of clone bundles. Besides, partial clone is better than full rollback and hooks running on initial clone are likely rare, so I think the impact is minimal.
2016-01-08 21:58:04 +03:00
$ cat server/access.log
clonebundles: support for seeding clones from pre-generated bundles Cloning can be an expensive operation for servers because the server generates a bundle from existing repository data at request time. For a large repository like mozilla-central, this consumes 4+ minutes of CPU time on the server. It also results in significant network utilization. Multiplied by hundreds or even thousands of clients and the ensuing load can result in difficulties scaling the Mercurial server. Despite generation of bundles being deterministic until the next changeset is added, the generation of bundles to service a clone request is not cached. Each clone thus performs redundant work. This is wasteful. This patch introduces the "clonebundles" extension and related client-side functionality to help alleviate this deficiency. The client-side feature is behind an experimental flag and is not enabled by default. It works as follows: 1) Server operator generates a bundle and makes it available on a server (likely HTTP). 2) Server operator defines the URL of a bundle file in a .hg/clonebundles.manifest file. 3) Client `hg clone`ing sees the server is advertising bundle URLs. 4) Client fetches and applies the advertised bundle. 5) Client performs equivalent of `hg pull` to fetch changes made since the bundle was created. Essentially, the server performs the expensive work of generating a bundle once and all subsequent clones fetch a static file from somewhere. Scaling static file serving is a much more manageable problem than scaling a Python application like Mercurial. Assuming your repository grows less than 1% per day, the end result is 99+% of CPU and network load from clones is eliminated, allowing Mercurial servers to scale more easily. Serving static files also means data can be transferred to clients as fast as they can consume it, rather than as fast as servers can generate it. This makes clones faster. Mozilla has implemented similar functionality of this patch on hg.mozilla.org using a custom extension. We are hosting bundle files in Amazon S3 and CloudFront (a CDN) and have successfully offloaded >1 TB/day in data transfer from hg.mozilla.org, freeing up significant bandwidth and CPU resources. The positive impact has been stellar and I believe it has proved its value to be included in Mercurial core. I feel it is important for the client-side support to be enabled in core by default because it means that clients will get faster, more reliable clones and will enable server operators to reduce load without requiring any client-side configuration changes (assuming clients are up to date, of course). The scope of this feature is narrowly and specifically tailored to cloning, despite "serve pulls from pre-generated bundles" being a valid and useful feature. I would eventually like for Mercurial servers to support transferring *all* repository data via statically hosted files. You could imagine a server that siphons all pushed data to bundle files and instructs clients to apply a stream of bundles to reconstruct all repository data. This feature, while useful and powerful, is significantly more work to implement because it requires the server component have awareness of discovery and a mapping of which changesets are in which files. Full, clone bundles, by contrast, are much simpler. The wire protocol command is named "clonebundles" instead of something more generic like "staticbundles" to leave the door open for a new, more powerful and more generic server-side component with minimal backwards compatibility implications. The name "bundleclone" is used by Mozilla's extension and would cause problems since there are subtle differences in Mozilla's extension. Mozilla's experience with this idea has taught us that some form of "content negotiation" is required. Not all clients will support all bundle formats or even URLs (advanced TLS requirements, etc). To ensure the highest uptake possible, a server needs to advertise multiple versions of bundles and clients need to be able to choose the most appropriate from that list one. The "attributes" in each server-advertised entry facilitate this filtering and sorting. Their use will become apparent in subsequent patches. Initial inspiration and credit for the idea of cloning from static files belongs to Augie Fackler and his "lookaside clone" extension proof of concept.
2015-10-09 21:22:01 +03:00
* - - [*] "GET /?cmd=capabilities HTTP/1.1" 200 - (glob)
* - - [*] "GET /?cmd=batch HTTP/1.1" 200 - x-hgarg-1:cmds=heads+%3Bknown+nodes%3D (glob)
bundle2: properly request phases during getbundle getbundle was requesting the "phase" namespace instead of the "phases" namespace, which led to the client still requesting the phases separately after getbundle finished.
2016-05-05 14:57:38 +03:00
* - - [*] "GET /?cmd=getbundle HTTP/1.1" 200 - x-hgarg-1:bundlecaps=HG20%2Cbundle2%3DHG20%250Achangegroup%253D01%252C02%250Adigests%253Dmd5%252Csha1%252Csha512%250Aerror%253Dabort%252Cunsupportedcontent%252Cpushraced%252Cpushkey%250Ahgtagsfnodes%250Alistkeys%250Apushkey%250Aremote-changegroup%253Dhttp%252Chttps&cg=1&common=0000000000000000000000000000000000000000&heads=aaff8d2ffbbf07a46dd1f05d8ae7877e3f56e2a2&listkeys=phases%2Cbookmarks (glob)
clonebundles: support for seeding clones from pre-generated bundles Cloning can be an expensive operation for servers because the server generates a bundle from existing repository data at request time. For a large repository like mozilla-central, this consumes 4+ minutes of CPU time on the server. It also results in significant network utilization. Multiplied by hundreds or even thousands of clients and the ensuing load can result in difficulties scaling the Mercurial server. Despite generation of bundles being deterministic until the next changeset is added, the generation of bundles to service a clone request is not cached. Each clone thus performs redundant work. This is wasteful. This patch introduces the "clonebundles" extension and related client-side functionality to help alleviate this deficiency. The client-side feature is behind an experimental flag and is not enabled by default. It works as follows: 1) Server operator generates a bundle and makes it available on a server (likely HTTP). 2) Server operator defines the URL of a bundle file in a .hg/clonebundles.manifest file. 3) Client `hg clone`ing sees the server is advertising bundle URLs. 4) Client fetches and applies the advertised bundle. 5) Client performs equivalent of `hg pull` to fetch changes made since the bundle was created. Essentially, the server performs the expensive work of generating a bundle once and all subsequent clones fetch a static file from somewhere. Scaling static file serving is a much more manageable problem than scaling a Python application like Mercurial. Assuming your repository grows less than 1% per day, the end result is 99+% of CPU and network load from clones is eliminated, allowing Mercurial servers to scale more easily. Serving static files also means data can be transferred to clients as fast as they can consume it, rather than as fast as servers can generate it. This makes clones faster. Mozilla has implemented similar functionality of this patch on hg.mozilla.org using a custom extension. We are hosting bundle files in Amazon S3 and CloudFront (a CDN) and have successfully offloaded >1 TB/day in data transfer from hg.mozilla.org, freeing up significant bandwidth and CPU resources. The positive impact has been stellar and I believe it has proved its value to be included in Mercurial core. I feel it is important for the client-side support to be enabled in core by default because it means that clients will get faster, more reliable clones and will enable server operators to reduce load without requiring any client-side configuration changes (assuming clients are up to date, of course). The scope of this feature is narrowly and specifically tailored to cloning, despite "serve pulls from pre-generated bundles" being a valid and useful feature. I would eventually like for Mercurial servers to support transferring *all* repository data via statically hosted files. You could imagine a server that siphons all pushed data to bundle files and instructs clients to apply a stream of bundles to reconstruct all repository data. This feature, while useful and powerful, is significantly more work to implement because it requires the server component have awareness of discovery and a mapping of which changesets are in which files. Full, clone bundles, by contrast, are much simpler. The wire protocol command is named "clonebundles" instead of something more generic like "staticbundles" to leave the door open for a new, more powerful and more generic server-side component with minimal backwards compatibility implications. The name "bundleclone" is used by Mozilla's extension and would cause problems since there are subtle differences in Mozilla's extension. Mozilla's experience with this idea has taught us that some form of "content negotiation" is required. Not all clients will support all bundle formats or even URLs (advanced TLS requirements, etc). To ensure the highest uptake possible, a server needs to advertise multiple versions of bundles and clients need to be able to choose the most appropriate from that list one. The "attributes" in each server-advertised entry facilitate this filtering and sorting. Their use will become apparent in subsequent patches. Initial inspiration and credit for the idea of cloning from static files belongs to Augie Fackler and his "lookaside clone" extension proof of concept.
2015-10-09 21:22:01 +03:00
Empty manifest file results in retrieval
(the extension only checks if the manifest file exists)
$ touch server/.hg/clonebundles.manifest
$ hg --verbose clone -U http://localhost:$HGPORT empty-manifest
no clone bundles available on remote; falling back to regular clone
requesting all changes
adding changesets
adding manifests
adding file changes
added 2 changesets with 2 changes to 2 files
Manifest file with invalid URL aborts
$ echo 'http://does.not.exist/bundle.hg' > server/.hg/clonebundles.manifest
$ hg clone http://localhost:$HGPORT 404-url
applying clone bundle from http://does.not.exist/bundle.hg
tests: handle getaddrinfo reporting "No address associated with hostname" This has been seen on some Fedora 23 systems.
2016-03-13 04:36:03 +03:00
error fetching bundle: (.* not known|getaddrinfo failed|No address associated with hostname) (re)
clonebundles: support for seeding clones from pre-generated bundles Cloning can be an expensive operation for servers because the server generates a bundle from existing repository data at request time. For a large repository like mozilla-central, this consumes 4+ minutes of CPU time on the server. It also results in significant network utilization. Multiplied by hundreds or even thousands of clients and the ensuing load can result in difficulties scaling the Mercurial server. Despite generation of bundles being deterministic until the next changeset is added, the generation of bundles to service a clone request is not cached. Each clone thus performs redundant work. This is wasteful. This patch introduces the "clonebundles" extension and related client-side functionality to help alleviate this deficiency. The client-side feature is behind an experimental flag and is not enabled by default. It works as follows: 1) Server operator generates a bundle and makes it available on a server (likely HTTP). 2) Server operator defines the URL of a bundle file in a .hg/clonebundles.manifest file. 3) Client `hg clone`ing sees the server is advertising bundle URLs. 4) Client fetches and applies the advertised bundle. 5) Client performs equivalent of `hg pull` to fetch changes made since the bundle was created. Essentially, the server performs the expensive work of generating a bundle once and all subsequent clones fetch a static file from somewhere. Scaling static file serving is a much more manageable problem than scaling a Python application like Mercurial. Assuming your repository grows less than 1% per day, the end result is 99+% of CPU and network load from clones is eliminated, allowing Mercurial servers to scale more easily. Serving static files also means data can be transferred to clients as fast as they can consume it, rather than as fast as servers can generate it. This makes clones faster. Mozilla has implemented similar functionality of this patch on hg.mozilla.org using a custom extension. We are hosting bundle files in Amazon S3 and CloudFront (a CDN) and have successfully offloaded >1 TB/day in data transfer from hg.mozilla.org, freeing up significant bandwidth and CPU resources. The positive impact has been stellar and I believe it has proved its value to be included in Mercurial core. I feel it is important for the client-side support to be enabled in core by default because it means that clients will get faster, more reliable clones and will enable server operators to reduce load without requiring any client-side configuration changes (assuming clients are up to date, of course). The scope of this feature is narrowly and specifically tailored to cloning, despite "serve pulls from pre-generated bundles" being a valid and useful feature. I would eventually like for Mercurial servers to support transferring *all* repository data via statically hosted files. You could imagine a server that siphons all pushed data to bundle files and instructs clients to apply a stream of bundles to reconstruct all repository data. This feature, while useful and powerful, is significantly more work to implement because it requires the server component have awareness of discovery and a mapping of which changesets are in which files. Full, clone bundles, by contrast, are much simpler. The wire protocol command is named "clonebundles" instead of something more generic like "staticbundles" to leave the door open for a new, more powerful and more generic server-side component with minimal backwards compatibility implications. The name "bundleclone" is used by Mozilla's extension and would cause problems since there are subtle differences in Mozilla's extension. Mozilla's experience with this idea has taught us that some form of "content negotiation" is required. Not all clients will support all bundle formats or even URLs (advanced TLS requirements, etc). To ensure the highest uptake possible, a server needs to advertise multiple versions of bundles and clients need to be able to choose the most appropriate from that list one. The "attributes" in each server-advertised entry facilitate this filtering and sorting. Their use will become apparent in subsequent patches. Initial inspiration and credit for the idea of cloning from static files belongs to Augie Fackler and his "lookaside clone" extension proof of concept.
2015-10-09 21:22:01 +03:00
abort: error applying bundle
exchange: make clone bundles non-experimental and enabled by default The clone bundles feature was introduced in Mercurial 3.6 behind an experimental and disabled by default flag. The feature has been enabled on hg.mozilla.org for a few months and has served many terabytes of clones. Users have been encouraged to use the feature and reception has been very positive (mainly due to faster clones as a result of connecting to a CDN). I have heard no feedback about changing the feature other than inquiries about when it will be enabled by default. So, I think the feature is ready to be enabled by default. This patch renames experimental.clonebundles to ui.clonebundles, documents the option, and enables it by default. References to the experimental state of clone bundles have been removed. The remaining config option docs in clonebundles.py have been removed because they are redudant with `hg help config`. There are some oddities with behavior of clone bundles. Because clones with clone bundles are effectively 2 `hg pull` operations, there may be 2 transactions. This could result in hooks running twice. If the subsequent pull is aborted, it could result in partial rollback and an incomplete clone. This behavior is a bit wonky and should probably be documented. If this patch is accepted, I'll send a follow-up to document it. I don't think this behavior should prevent the feature being enabled by default. Reworking the clone mechanism to support interrupted or multi-part clones feels like a major new feature and something that when implemented can change the hook and rollback semantics of clone bundles. Besides, partial clone is better than full rollback and hooks running on initial clone are likely rare, so I think the impact is minimal.
2016-01-08 21:58:04 +03:00
(if this error persists, consider contacting the server operator or disable clone bundles via "--config ui.clonebundles=false")
clonebundles: support for seeding clones from pre-generated bundles Cloning can be an expensive operation for servers because the server generates a bundle from existing repository data at request time. For a large repository like mozilla-central, this consumes 4+ minutes of CPU time on the server. It also results in significant network utilization. Multiplied by hundreds or even thousands of clients and the ensuing load can result in difficulties scaling the Mercurial server. Despite generation of bundles being deterministic until the next changeset is added, the generation of bundles to service a clone request is not cached. Each clone thus performs redundant work. This is wasteful. This patch introduces the "clonebundles" extension and related client-side functionality to help alleviate this deficiency. The client-side feature is behind an experimental flag and is not enabled by default. It works as follows: 1) Server operator generates a bundle and makes it available on a server (likely HTTP). 2) Server operator defines the URL of a bundle file in a .hg/clonebundles.manifest file. 3) Client `hg clone`ing sees the server is advertising bundle URLs. 4) Client fetches and applies the advertised bundle. 5) Client performs equivalent of `hg pull` to fetch changes made since the bundle was created. Essentially, the server performs the expensive work of generating a bundle once and all subsequent clones fetch a static file from somewhere. Scaling static file serving is a much more manageable problem than scaling a Python application like Mercurial. Assuming your repository grows less than 1% per day, the end result is 99+% of CPU and network load from clones is eliminated, allowing Mercurial servers to scale more easily. Serving static files also means data can be transferred to clients as fast as they can consume it, rather than as fast as servers can generate it. This makes clones faster. Mozilla has implemented similar functionality of this patch on hg.mozilla.org using a custom extension. We are hosting bundle files in Amazon S3 and CloudFront (a CDN) and have successfully offloaded >1 TB/day in data transfer from hg.mozilla.org, freeing up significant bandwidth and CPU resources. The positive impact has been stellar and I believe it has proved its value to be included in Mercurial core. I feel it is important for the client-side support to be enabled in core by default because it means that clients will get faster, more reliable clones and will enable server operators to reduce load without requiring any client-side configuration changes (assuming clients are up to date, of course). The scope of this feature is narrowly and specifically tailored to cloning, despite "serve pulls from pre-generated bundles" being a valid and useful feature. I would eventually like for Mercurial servers to support transferring *all* repository data via statically hosted files. You could imagine a server that siphons all pushed data to bundle files and instructs clients to apply a stream of bundles to reconstruct all repository data. This feature, while useful and powerful, is significantly more work to implement because it requires the server component have awareness of discovery and a mapping of which changesets are in which files. Full, clone bundles, by contrast, are much simpler. The wire protocol command is named "clonebundles" instead of something more generic like "staticbundles" to leave the door open for a new, more powerful and more generic server-side component with minimal backwards compatibility implications. The name "bundleclone" is used by Mozilla's extension and would cause problems since there are subtle differences in Mozilla's extension. Mozilla's experience with this idea has taught us that some form of "content negotiation" is required. Not all clients will support all bundle formats or even URLs (advanced TLS requirements, etc). To ensure the highest uptake possible, a server needs to advertise multiple versions of bundles and clients need to be able to choose the most appropriate from that list one. The "attributes" in each server-advertised entry facilitate this filtering and sorting. Their use will become apparent in subsequent patches. Initial inspiration and credit for the idea of cloning from static files belongs to Augie Fackler and his "lookaside clone" extension proof of concept.
2015-10-09 21:22:01 +03:00
[255]
Server is not running aborts
$ echo "http://localhost:$HGPORT1/bundle.hg" > server/.hg/clonebundles.manifest
$ hg clone http://localhost:$HGPORT server-not-runner
applying clone bundle from http://localhost:$HGPORT1/bundle.hg
tests: tolerate differences between Linux and Windows error strings These are related to differences in how missing files and network connection failures are displayed. I opted to combine the strings in one line instead of using '#if windows' blocks around entire commands in order to avoid future changes being accidentally missed in the Windows sections. Globbing away the entire output seemed wrong, as it could mask other failures. The raw messages involved are: Linux Windows "* not known" <-> "getaddrinfo failed" "Connection refused" <-> "No connection could be made because the target machine actively refused it" "No such file or directory" <-> "The system cannot find the file specified" Issue 4941 indicates that NetBSD has yet another string for "* not known". Also, the histedit test shows that the missing file is printed first on Windows, last on Linux. That is controlled in windows.py:posixfile if we care to change it.
2015-11-09 21:12:35 +03:00
error fetching bundle: * refused* (glob)
clonebundles: support for seeding clones from pre-generated bundles Cloning can be an expensive operation for servers because the server generates a bundle from existing repository data at request time. For a large repository like mozilla-central, this consumes 4+ minutes of CPU time on the server. It also results in significant network utilization. Multiplied by hundreds or even thousands of clients and the ensuing load can result in difficulties scaling the Mercurial server. Despite generation of bundles being deterministic until the next changeset is added, the generation of bundles to service a clone request is not cached. Each clone thus performs redundant work. This is wasteful. This patch introduces the "clonebundles" extension and related client-side functionality to help alleviate this deficiency. The client-side feature is behind an experimental flag and is not enabled by default. It works as follows: 1) Server operator generates a bundle and makes it available on a server (likely HTTP). 2) Server operator defines the URL of a bundle file in a .hg/clonebundles.manifest file. 3) Client `hg clone`ing sees the server is advertising bundle URLs. 4) Client fetches and applies the advertised bundle. 5) Client performs equivalent of `hg pull` to fetch changes made since the bundle was created. Essentially, the server performs the expensive work of generating a bundle once and all subsequent clones fetch a static file from somewhere. Scaling static file serving is a much more manageable problem than scaling a Python application like Mercurial. Assuming your repository grows less than 1% per day, the end result is 99+% of CPU and network load from clones is eliminated, allowing Mercurial servers to scale more easily. Serving static files also means data can be transferred to clients as fast as they can consume it, rather than as fast as servers can generate it. This makes clones faster. Mozilla has implemented similar functionality of this patch on hg.mozilla.org using a custom extension. We are hosting bundle files in Amazon S3 and CloudFront (a CDN) and have successfully offloaded >1 TB/day in data transfer from hg.mozilla.org, freeing up significant bandwidth and CPU resources. The positive impact has been stellar and I believe it has proved its value to be included in Mercurial core. I feel it is important for the client-side support to be enabled in core by default because it means that clients will get faster, more reliable clones and will enable server operators to reduce load without requiring any client-side configuration changes (assuming clients are up to date, of course). The scope of this feature is narrowly and specifically tailored to cloning, despite "serve pulls from pre-generated bundles" being a valid and useful feature. I would eventually like for Mercurial servers to support transferring *all* repository data via statically hosted files. You could imagine a server that siphons all pushed data to bundle files and instructs clients to apply a stream of bundles to reconstruct all repository data. This feature, while useful and powerful, is significantly more work to implement because it requires the server component have awareness of discovery and a mapping of which changesets are in which files. Full, clone bundles, by contrast, are much simpler. The wire protocol command is named "clonebundles" instead of something more generic like "staticbundles" to leave the door open for a new, more powerful and more generic server-side component with minimal backwards compatibility implications. The name "bundleclone" is used by Mozilla's extension and would cause problems since there are subtle differences in Mozilla's extension. Mozilla's experience with this idea has taught us that some form of "content negotiation" is required. Not all clients will support all bundle formats or even URLs (advanced TLS requirements, etc). To ensure the highest uptake possible, a server needs to advertise multiple versions of bundles and clients need to be able to choose the most appropriate from that list one. The "attributes" in each server-advertised entry facilitate this filtering and sorting. Their use will become apparent in subsequent patches. Initial inspiration and credit for the idea of cloning from static files belongs to Augie Fackler and his "lookaside clone" extension proof of concept.
2015-10-09 21:22:01 +03:00
abort: error applying bundle
exchange: make clone bundles non-experimental and enabled by default The clone bundles feature was introduced in Mercurial 3.6 behind an experimental and disabled by default flag. The feature has been enabled on hg.mozilla.org for a few months and has served many terabytes of clones. Users have been encouraged to use the feature and reception has been very positive (mainly due to faster clones as a result of connecting to a CDN). I have heard no feedback about changing the feature other than inquiries about when it will be enabled by default. So, I think the feature is ready to be enabled by default. This patch renames experimental.clonebundles to ui.clonebundles, documents the option, and enables it by default. References to the experimental state of clone bundles have been removed. The remaining config option docs in clonebundles.py have been removed because they are redudant with `hg help config`. There are some oddities with behavior of clone bundles. Because clones with clone bundles are effectively 2 `hg pull` operations, there may be 2 transactions. This could result in hooks running twice. If the subsequent pull is aborted, it could result in partial rollback and an incomplete clone. This behavior is a bit wonky and should probably be documented. If this patch is accepted, I'll send a follow-up to document it. I don't think this behavior should prevent the feature being enabled by default. Reworking the clone mechanism to support interrupted or multi-part clones feels like a major new feature and something that when implemented can change the hook and rollback semantics of clone bundles. Besides, partial clone is better than full rollback and hooks running on initial clone are likely rare, so I think the impact is minimal.
2016-01-08 21:58:04 +03:00
(if this error persists, consider contacting the server operator or disable clone bundles via "--config ui.clonebundles=false")
clonebundles: support for seeding clones from pre-generated bundles Cloning can be an expensive operation for servers because the server generates a bundle from existing repository data at request time. For a large repository like mozilla-central, this consumes 4+ minutes of CPU time on the server. It also results in significant network utilization. Multiplied by hundreds or even thousands of clients and the ensuing load can result in difficulties scaling the Mercurial server. Despite generation of bundles being deterministic until the next changeset is added, the generation of bundles to service a clone request is not cached. Each clone thus performs redundant work. This is wasteful. This patch introduces the "clonebundles" extension and related client-side functionality to help alleviate this deficiency. The client-side feature is behind an experimental flag and is not enabled by default. It works as follows: 1) Server operator generates a bundle and makes it available on a server (likely HTTP). 2) Server operator defines the URL of a bundle file in a .hg/clonebundles.manifest file. 3) Client `hg clone`ing sees the server is advertising bundle URLs. 4) Client fetches and applies the advertised bundle. 5) Client performs equivalent of `hg pull` to fetch changes made since the bundle was created. Essentially, the server performs the expensive work of generating a bundle once and all subsequent clones fetch a static file from somewhere. Scaling static file serving is a much more manageable problem than scaling a Python application like Mercurial. Assuming your repository grows less than 1% per day, the end result is 99+% of CPU and network load from clones is eliminated, allowing Mercurial servers to scale more easily. Serving static files also means data can be transferred to clients as fast as they can consume it, rather than as fast as servers can generate it. This makes clones faster. Mozilla has implemented similar functionality of this patch on hg.mozilla.org using a custom extension. We are hosting bundle files in Amazon S3 and CloudFront (a CDN) and have successfully offloaded >1 TB/day in data transfer from hg.mozilla.org, freeing up significant bandwidth and CPU resources. The positive impact has been stellar and I believe it has proved its value to be included in Mercurial core. I feel it is important for the client-side support to be enabled in core by default because it means that clients will get faster, more reliable clones and will enable server operators to reduce load without requiring any client-side configuration changes (assuming clients are up to date, of course). The scope of this feature is narrowly and specifically tailored to cloning, despite "serve pulls from pre-generated bundles" being a valid and useful feature. I would eventually like for Mercurial servers to support transferring *all* repository data via statically hosted files. You could imagine a server that siphons all pushed data to bundle files and instructs clients to apply a stream of bundles to reconstruct all repository data. This feature, while useful and powerful, is significantly more work to implement because it requires the server component have awareness of discovery and a mapping of which changesets are in which files. Full, clone bundles, by contrast, are much simpler. The wire protocol command is named "clonebundles" instead of something more generic like "staticbundles" to leave the door open for a new, more powerful and more generic server-side component with minimal backwards compatibility implications. The name "bundleclone" is used by Mozilla's extension and would cause problems since there are subtle differences in Mozilla's extension. Mozilla's experience with this idea has taught us that some form of "content negotiation" is required. Not all clients will support all bundle formats or even URLs (advanced TLS requirements, etc). To ensure the highest uptake possible, a server needs to advertise multiple versions of bundles and clients need to be able to choose the most appropriate from that list one. The "attributes" in each server-advertised entry facilitate this filtering and sorting. Their use will become apparent in subsequent patches. Initial inspiration and credit for the idea of cloning from static files belongs to Augie Fackler and his "lookaside clone" extension proof of concept.
2015-10-09 21:22:01 +03:00
[255]
Server returns 404
$ python $TESTDIR/dumbhttp.py -p $HGPORT1 --pid http.pid
$ cat http.pid >> $DAEMON_PIDS
$ hg clone http://localhost:$HGPORT running-404
applying clone bundle from http://localhost:$HGPORT1/bundle.hg
HTTP error fetching bundle: HTTP Error 404: File not found
abort: error applying bundle
exchange: make clone bundles non-experimental and enabled by default The clone bundles feature was introduced in Mercurial 3.6 behind an experimental and disabled by default flag. The feature has been enabled on hg.mozilla.org for a few months and has served many terabytes of clones. Users have been encouraged to use the feature and reception has been very positive (mainly due to faster clones as a result of connecting to a CDN). I have heard no feedback about changing the feature other than inquiries about when it will be enabled by default. So, I think the feature is ready to be enabled by default. This patch renames experimental.clonebundles to ui.clonebundles, documents the option, and enables it by default. References to the experimental state of clone bundles have been removed. The remaining config option docs in clonebundles.py have been removed because they are redudant with `hg help config`. There are some oddities with behavior of clone bundles. Because clones with clone bundles are effectively 2 `hg pull` operations, there may be 2 transactions. This could result in hooks running twice. If the subsequent pull is aborted, it could result in partial rollback and an incomplete clone. This behavior is a bit wonky and should probably be documented. If this patch is accepted, I'll send a follow-up to document it. I don't think this behavior should prevent the feature being enabled by default. Reworking the clone mechanism to support interrupted or multi-part clones feels like a major new feature and something that when implemented can change the hook and rollback semantics of clone bundles. Besides, partial clone is better than full rollback and hooks running on initial clone are likely rare, so I think the impact is minimal.
2016-01-08 21:58:04 +03:00
(if this error persists, consider contacting the server operator or disable clone bundles via "--config ui.clonebundles=false")
clonebundles: support for seeding clones from pre-generated bundles Cloning can be an expensive operation for servers because the server generates a bundle from existing repository data at request time. For a large repository like mozilla-central, this consumes 4+ minutes of CPU time on the server. It also results in significant network utilization. Multiplied by hundreds or even thousands of clients and the ensuing load can result in difficulties scaling the Mercurial server. Despite generation of bundles being deterministic until the next changeset is added, the generation of bundles to service a clone request is not cached. Each clone thus performs redundant work. This is wasteful. This patch introduces the "clonebundles" extension and related client-side functionality to help alleviate this deficiency. The client-side feature is behind an experimental flag and is not enabled by default. It works as follows: 1) Server operator generates a bundle and makes it available on a server (likely HTTP). 2) Server operator defines the URL of a bundle file in a .hg/clonebundles.manifest file. 3) Client `hg clone`ing sees the server is advertising bundle URLs. 4) Client fetches and applies the advertised bundle. 5) Client performs equivalent of `hg pull` to fetch changes made since the bundle was created. Essentially, the server performs the expensive work of generating a bundle once and all subsequent clones fetch a static file from somewhere. Scaling static file serving is a much more manageable problem than scaling a Python application like Mercurial. Assuming your repository grows less than 1% per day, the end result is 99+% of CPU and network load from clones is eliminated, allowing Mercurial servers to scale more easily. Serving static files also means data can be transferred to clients as fast as they can consume it, rather than as fast as servers can generate it. This makes clones faster. Mozilla has implemented similar functionality of this patch on hg.mozilla.org using a custom extension. We are hosting bundle files in Amazon S3 and CloudFront (a CDN) and have successfully offloaded >1 TB/day in data transfer from hg.mozilla.org, freeing up significant bandwidth and CPU resources. The positive impact has been stellar and I believe it has proved its value to be included in Mercurial core. I feel it is important for the client-side support to be enabled in core by default because it means that clients will get faster, more reliable clones and will enable server operators to reduce load without requiring any client-side configuration changes (assuming clients are up to date, of course). The scope of this feature is narrowly and specifically tailored to cloning, despite "serve pulls from pre-generated bundles" being a valid and useful feature. I would eventually like for Mercurial servers to support transferring *all* repository data via statically hosted files. You could imagine a server that siphons all pushed data to bundle files and instructs clients to apply a stream of bundles to reconstruct all repository data. This feature, while useful and powerful, is significantly more work to implement because it requires the server component have awareness of discovery and a mapping of which changesets are in which files. Full, clone bundles, by contrast, are much simpler. The wire protocol command is named "clonebundles" instead of something more generic like "staticbundles" to leave the door open for a new, more powerful and more generic server-side component with minimal backwards compatibility implications. The name "bundleclone" is used by Mozilla's extension and would cause problems since there are subtle differences in Mozilla's extension. Mozilla's experience with this idea has taught us that some form of "content negotiation" is required. Not all clients will support all bundle formats or even URLs (advanced TLS requirements, etc). To ensure the highest uptake possible, a server needs to advertise multiple versions of bundles and clients need to be able to choose the most appropriate from that list one. The "attributes" in each server-advertised entry facilitate this filtering and sorting. Their use will become apparent in subsequent patches. Initial inspiration and credit for the idea of cloning from static files belongs to Augie Fackler and his "lookaside clone" extension proof of concept.
2015-10-09 21:22:01 +03:00
[255]
We can override failure to fall back to regular clone
$ hg --config ui.clonebundlefallback=true clone -U http://localhost:$HGPORT 404-fallback
applying clone bundle from http://localhost:$HGPORT1/bundle.hg
HTTP error fetching bundle: HTTP Error 404: File not found
falling back to normal clone
requesting all changes
adding changesets
adding manifests
adding file changes
added 2 changesets with 2 changes to 2 files
Bundle with partial content works
clonebundles: filter on bundle specification Not all clients are capable of reading every bundle. Currently, content negotiation to ensure a server sends a client a compatible bundle format is performed at request time. The response bundle is dynamically generated at request time, so this works fine. Clone bundles are statically generated *before* the request. This means that a modern server could produce bundles that a legacy client isn't capable of reading. Without some kind of "type hint" in the clone bundles manifest, a client may attempt to download an incompatible bundle. Furthermore, a client may not realize a bundle is incompatible until it has processed part of the bundle (imagine consuming a 1 GB changegroup bundle2 part only to discover the bundle2 part afterwards is incompatibl). This would waste time and resources. And it isn't very user friendly. Clone bundle manifests thus need to advertise the *exact* format of the hosted bundles so clients may filter out entries that they don't know how to read. This patch introduces that mechanism. We introduce the BUNDLESPEC attribute to declare the "bundle specification" of the entry. Bundle specifications are parsed using exchange.parsebundlespecification, which uses the same strings as the "--type" argument to `hg bundle`. The supported bundle specifications are well defined and backwards compatible. When a client encounters a BUNDLESPEC that is invalid or unsupported, it silently ignores the entry.
2015-10-13 21:45:30 +03:00
$ hg -R server bundle --type gzip-v1 --base null -r 53245c60e682 partial.hg
clonebundles: support for seeding clones from pre-generated bundles Cloning can be an expensive operation for servers because the server generates a bundle from existing repository data at request time. For a large repository like mozilla-central, this consumes 4+ minutes of CPU time on the server. It also results in significant network utilization. Multiplied by hundreds or even thousands of clients and the ensuing load can result in difficulties scaling the Mercurial server. Despite generation of bundles being deterministic until the next changeset is added, the generation of bundles to service a clone request is not cached. Each clone thus performs redundant work. This is wasteful. This patch introduces the "clonebundles" extension and related client-side functionality to help alleviate this deficiency. The client-side feature is behind an experimental flag and is not enabled by default. It works as follows: 1) Server operator generates a bundle and makes it available on a server (likely HTTP). 2) Server operator defines the URL of a bundle file in a .hg/clonebundles.manifest file. 3) Client `hg clone`ing sees the server is advertising bundle URLs. 4) Client fetches and applies the advertised bundle. 5) Client performs equivalent of `hg pull` to fetch changes made since the bundle was created. Essentially, the server performs the expensive work of generating a bundle once and all subsequent clones fetch a static file from somewhere. Scaling static file serving is a much more manageable problem than scaling a Python application like Mercurial. Assuming your repository grows less than 1% per day, the end result is 99+% of CPU and network load from clones is eliminated, allowing Mercurial servers to scale more easily. Serving static files also means data can be transferred to clients as fast as they can consume it, rather than as fast as servers can generate it. This makes clones faster. Mozilla has implemented similar functionality of this patch on hg.mozilla.org using a custom extension. We are hosting bundle files in Amazon S3 and CloudFront (a CDN) and have successfully offloaded >1 TB/day in data transfer from hg.mozilla.org, freeing up significant bandwidth and CPU resources. The positive impact has been stellar and I believe it has proved its value to be included in Mercurial core. I feel it is important for the client-side support to be enabled in core by default because it means that clients will get faster, more reliable clones and will enable server operators to reduce load without requiring any client-side configuration changes (assuming clients are up to date, of course). The scope of this feature is narrowly and specifically tailored to cloning, despite "serve pulls from pre-generated bundles" being a valid and useful feature. I would eventually like for Mercurial servers to support transferring *all* repository data via statically hosted files. You could imagine a server that siphons all pushed data to bundle files and instructs clients to apply a stream of bundles to reconstruct all repository data. This feature, while useful and powerful, is significantly more work to implement because it requires the server component have awareness of discovery and a mapping of which changesets are in which files. Full, clone bundles, by contrast, are much simpler. The wire protocol command is named "clonebundles" instead of something more generic like "staticbundles" to leave the door open for a new, more powerful and more generic server-side component with minimal backwards compatibility implications. The name "bundleclone" is used by Mozilla's extension and would cause problems since there are subtle differences in Mozilla's extension. Mozilla's experience with this idea has taught us that some form of "content negotiation" is required. Not all clients will support all bundle formats or even URLs (advanced TLS requirements, etc). To ensure the highest uptake possible, a server needs to advertise multiple versions of bundles and clients need to be able to choose the most appropriate from that list one. The "attributes" in each server-advertised entry facilitate this filtering and sorting. Their use will become apparent in subsequent patches. Initial inspiration and credit for the idea of cloning from static files belongs to Augie Fackler and his "lookaside clone" extension proof of concept.
2015-10-09 21:22:01 +03:00
1 changesets found
clonebundles: filter on bundle specification Not all clients are capable of reading every bundle. Currently, content negotiation to ensure a server sends a client a compatible bundle format is performed at request time. The response bundle is dynamically generated at request time, so this works fine. Clone bundles are statically generated *before* the request. This means that a modern server could produce bundles that a legacy client isn't capable of reading. Without some kind of "type hint" in the clone bundles manifest, a client may attempt to download an incompatible bundle. Furthermore, a client may not realize a bundle is incompatible until it has processed part of the bundle (imagine consuming a 1 GB changegroup bundle2 part only to discover the bundle2 part afterwards is incompatibl). This would waste time and resources. And it isn't very user friendly. Clone bundle manifests thus need to advertise the *exact* format of the hosted bundles so clients may filter out entries that they don't know how to read. This patch introduces that mechanism. We introduce the BUNDLESPEC attribute to declare the "bundle specification" of the entry. Bundle specifications are parsed using exchange.parsebundlespecification, which uses the same strings as the "--type" argument to `hg bundle`. The supported bundle specifications are well defined and backwards compatible. When a client encounters a BUNDLESPEC that is invalid or unsupported, it silently ignores the entry.
2015-10-13 21:45:30 +03:00
We verify exact bundle content as an extra check against accidental future
changes. If this output changes, we could break old clients.
$ f --size --hexdump partial.hg
mdiff: don't emit a diff header for empty trivial deltas An empty trivial delta, coded as (0, 0, 0) makes the delta application do nothing, but still takes 12 bytes, while skipping it altogether works as much, without taking any space at all.
2016-01-12 06:00:07 +03:00
partial.hg: size=207
clonebundles: filter on bundle specification Not all clients are capable of reading every bundle. Currently, content negotiation to ensure a server sends a client a compatible bundle format is performed at request time. The response bundle is dynamically generated at request time, so this works fine. Clone bundles are statically generated *before* the request. This means that a modern server could produce bundles that a legacy client isn't capable of reading. Without some kind of "type hint" in the clone bundles manifest, a client may attempt to download an incompatible bundle. Furthermore, a client may not realize a bundle is incompatible until it has processed part of the bundle (imagine consuming a 1 GB changegroup bundle2 part only to discover the bundle2 part afterwards is incompatibl). This would waste time and resources. And it isn't very user friendly. Clone bundle manifests thus need to advertise the *exact* format of the hosted bundles so clients may filter out entries that they don't know how to read. This patch introduces that mechanism. We introduce the BUNDLESPEC attribute to declare the "bundle specification" of the entry. Bundle specifications are parsed using exchange.parsebundlespecification, which uses the same strings as the "--type" argument to `hg bundle`. The supported bundle specifications are well defined and backwards compatible. When a client encounters a BUNDLESPEC that is invalid or unsupported, it silently ignores the entry.
2015-10-13 21:45:30 +03:00
0000: 48 47 31 30 47 5a 78 9c 63 60 60 98 17 ac 12 93 |HG10GZx.c``.....|
0010: f0 ac a9 23 45 70 cb bf 0d 5f 59 4e 4a 7f 79 21 |...#Ep..._YNJ.y!|
0020: 9b cc 40 24 20 a0 d7 ce 2c d1 38 25 cd 24 25 d5 |..@$ ...,.8%.$%.|
0030: d8 c2 22 cd 38 d9 24 cd 22 d5 c8 22 cd 24 cd 32 |..".8.$."..".$.2|
0040: d1 c2 d0 c4 c8 d2 32 d1 38 39 29 c9 34 cd d4 80 |......2.89).4...|
0050: ab 24 b5 b8 84 cb 40 c1 80 2b 2d 3f 9f 8b 2b 31 |.$....@..+-?..+1|
0060: 25 45 01 c8 80 9a d2 9b 65 fb e5 9e 45 bf 8d 7f |%E......e...E...|
0070: 9f c6 97 9f 2b 44 34 67 d9 ec 8e 0f a0 92 0b 75 |....+D4g.......u|
0080: 41 d6 24 59 18 a4 a4 9a a6 18 1a 5b 98 9b 5a 98 |A.$Y.......[..Z.|
0090: 9a 18 26 9b a6 19 98 1a 99 99 26 a6 18 9a 98 24 |..&.......&....$|
00a0: 26 59 a6 25 5a 98 a5 18 a6 24 71 41 35 b1 43 dc |&Y.%Z....$qA5.C.|
mdiff: don't emit a diff header for empty trivial deltas An empty trivial delta, coded as (0, 0, 0) makes the delta application do nothing, but still takes 12 bytes, while skipping it altogether works as much, without taking any space at all.
2016-01-12 06:00:07 +03:00
00b0: 16 b2 83 f7 e9 45 8b d2 56 c7 a3 1f 82 52 d7 8a |.....E..V....R..|
00c0: 78 ed fc d5 76 f1 36 35 dc 05 00 36 ed 5e c7 |x...v.65...6.^.|
clonebundles: filter on bundle specification Not all clients are capable of reading every bundle. Currently, content negotiation to ensure a server sends a client a compatible bundle format is performed at request time. The response bundle is dynamically generated at request time, so this works fine. Clone bundles are statically generated *before* the request. This means that a modern server could produce bundles that a legacy client isn't capable of reading. Without some kind of "type hint" in the clone bundles manifest, a client may attempt to download an incompatible bundle. Furthermore, a client may not realize a bundle is incompatible until it has processed part of the bundle (imagine consuming a 1 GB changegroup bundle2 part only to discover the bundle2 part afterwards is incompatibl). This would waste time and resources. And it isn't very user friendly. Clone bundle manifests thus need to advertise the *exact* format of the hosted bundles so clients may filter out entries that they don't know how to read. This patch introduces that mechanism. We introduce the BUNDLESPEC attribute to declare the "bundle specification" of the entry. Bundle specifications are parsed using exchange.parsebundlespecification, which uses the same strings as the "--type" argument to `hg bundle`. The supported bundle specifications are well defined and backwards compatible. When a client encounters a BUNDLESPEC that is invalid or unsupported, it silently ignores the entry.
2015-10-13 21:45:30 +03:00
clonebundles: support for seeding clones from pre-generated bundles Cloning can be an expensive operation for servers because the server generates a bundle from existing repository data at request time. For a large repository like mozilla-central, this consumes 4+ minutes of CPU time on the server. It also results in significant network utilization. Multiplied by hundreds or even thousands of clients and the ensuing load can result in difficulties scaling the Mercurial server. Despite generation of bundles being deterministic until the next changeset is added, the generation of bundles to service a clone request is not cached. Each clone thus performs redundant work. This is wasteful. This patch introduces the "clonebundles" extension and related client-side functionality to help alleviate this deficiency. The client-side feature is behind an experimental flag and is not enabled by default. It works as follows: 1) Server operator generates a bundle and makes it available on a server (likely HTTP). 2) Server operator defines the URL of a bundle file in a .hg/clonebundles.manifest file. 3) Client `hg clone`ing sees the server is advertising bundle URLs. 4) Client fetches and applies the advertised bundle. 5) Client performs equivalent of `hg pull` to fetch changes made since the bundle was created. Essentially, the server performs the expensive work of generating a bundle once and all subsequent clones fetch a static file from somewhere. Scaling static file serving is a much more manageable problem than scaling a Python application like Mercurial. Assuming your repository grows less than 1% per day, the end result is 99+% of CPU and network load from clones is eliminated, allowing Mercurial servers to scale more easily. Serving static files also means data can be transferred to clients as fast as they can consume it, rather than as fast as servers can generate it. This makes clones faster. Mozilla has implemented similar functionality of this patch on hg.mozilla.org using a custom extension. We are hosting bundle files in Amazon S3 and CloudFront (a CDN) and have successfully offloaded >1 TB/day in data transfer from hg.mozilla.org, freeing up significant bandwidth and CPU resources. The positive impact has been stellar and I believe it has proved its value to be included in Mercurial core. I feel it is important for the client-side support to be enabled in core by default because it means that clients will get faster, more reliable clones and will enable server operators to reduce load without requiring any client-side configuration changes (assuming clients are up to date, of course). The scope of this feature is narrowly and specifically tailored to cloning, despite "serve pulls from pre-generated bundles" being a valid and useful feature. I would eventually like for Mercurial servers to support transferring *all* repository data via statically hosted files. You could imagine a server that siphons all pushed data to bundle files and instructs clients to apply a stream of bundles to reconstruct all repository data. This feature, while useful and powerful, is significantly more work to implement because it requires the server component have awareness of discovery and a mapping of which changesets are in which files. Full, clone bundles, by contrast, are much simpler. The wire protocol command is named "clonebundles" instead of something more generic like "staticbundles" to leave the door open for a new, more powerful and more generic server-side component with minimal backwards compatibility implications. The name "bundleclone" is used by Mozilla's extension and would cause problems since there are subtle differences in Mozilla's extension. Mozilla's experience with this idea has taught us that some form of "content negotiation" is required. Not all clients will support all bundle formats or even URLs (advanced TLS requirements, etc). To ensure the highest uptake possible, a server needs to advertise multiple versions of bundles and clients need to be able to choose the most appropriate from that list one. The "attributes" in each server-advertised entry facilitate this filtering and sorting. Their use will become apparent in subsequent patches. Initial inspiration and credit for the idea of cloning from static files belongs to Augie Fackler and his "lookaside clone" extension proof of concept.
2015-10-09 21:22:01 +03:00
$ echo "http://localhost:$HGPORT1/partial.hg" > server/.hg/clonebundles.manifest
$ hg clone -U http://localhost:$HGPORT partial-bundle
applying clone bundle from http://localhost:$HGPORT1/partial.hg
adding changesets
adding manifests
adding file changes
added 1 changesets with 1 changes to 1 files
finished applying clone bundle
searching for changes
adding changesets
adding manifests
adding file changes
added 1 changesets with 1 changes to 1 files
test-clonebundles.t: add test for incremental pull This demonstrates issue4932.
2015-11-03 23:15:14 +03:00
Incremental pull doesn't fetch bundle
$ hg clone -r 53245c60e682 -U http://localhost:$HGPORT partial-clone
adding changesets
adding manifests
adding file changes
added 1 changesets with 1 changes to 1 files
$ cd partial-clone
$ hg pull
pulling from http://localhost:$HGPORT/
searching for changes
adding changesets
adding manifests
adding file changes
added 1 changesets with 1 changes to 1 files
(run 'hg update' to get a working copy)
$ cd ..
clonebundles: support for seeding clones from pre-generated bundles Cloning can be an expensive operation for servers because the server generates a bundle from existing repository data at request time. For a large repository like mozilla-central, this consumes 4+ minutes of CPU time on the server. It also results in significant network utilization. Multiplied by hundreds or even thousands of clients and the ensuing load can result in difficulties scaling the Mercurial server. Despite generation of bundles being deterministic until the next changeset is added, the generation of bundles to service a clone request is not cached. Each clone thus performs redundant work. This is wasteful. This patch introduces the "clonebundles" extension and related client-side functionality to help alleviate this deficiency. The client-side feature is behind an experimental flag and is not enabled by default. It works as follows: 1) Server operator generates a bundle and makes it available on a server (likely HTTP). 2) Server operator defines the URL of a bundle file in a .hg/clonebundles.manifest file. 3) Client `hg clone`ing sees the server is advertising bundle URLs. 4) Client fetches and applies the advertised bundle. 5) Client performs equivalent of `hg pull` to fetch changes made since the bundle was created. Essentially, the server performs the expensive work of generating a bundle once and all subsequent clones fetch a static file from somewhere. Scaling static file serving is a much more manageable problem than scaling a Python application like Mercurial. Assuming your repository grows less than 1% per day, the end result is 99+% of CPU and network load from clones is eliminated, allowing Mercurial servers to scale more easily. Serving static files also means data can be transferred to clients as fast as they can consume it, rather than as fast as servers can generate it. This makes clones faster. Mozilla has implemented similar functionality of this patch on hg.mozilla.org using a custom extension. We are hosting bundle files in Amazon S3 and CloudFront (a CDN) and have successfully offloaded >1 TB/day in data transfer from hg.mozilla.org, freeing up significant bandwidth and CPU resources. The positive impact has been stellar and I believe it has proved its value to be included in Mercurial core. I feel it is important for the client-side support to be enabled in core by default because it means that clients will get faster, more reliable clones and will enable server operators to reduce load without requiring any client-side configuration changes (assuming clients are up to date, of course). The scope of this feature is narrowly and specifically tailored to cloning, despite "serve pulls from pre-generated bundles" being a valid and useful feature. I would eventually like for Mercurial servers to support transferring *all* repository data via statically hosted files. You could imagine a server that siphons all pushed data to bundle files and instructs clients to apply a stream of bundles to reconstruct all repository data. This feature, while useful and powerful, is significantly more work to implement because it requires the server component have awareness of discovery and a mapping of which changesets are in which files. Full, clone bundles, by contrast, are much simpler. The wire protocol command is named "clonebundles" instead of something more generic like "staticbundles" to leave the door open for a new, more powerful and more generic server-side component with minimal backwards compatibility implications. The name "bundleclone" is used by Mozilla's extension and would cause problems since there are subtle differences in Mozilla's extension. Mozilla's experience with this idea has taught us that some form of "content negotiation" is required. Not all clients will support all bundle formats or even URLs (advanced TLS requirements, etc). To ensure the highest uptake possible, a server needs to advertise multiple versions of bundles and clients need to be able to choose the most appropriate from that list one. The "attributes" in each server-advertised entry facilitate this filtering and sorting. Their use will become apparent in subsequent patches. Initial inspiration and credit for the idea of cloning from static files belongs to Augie Fackler and his "lookaside clone" extension proof of concept.
2015-10-09 21:22:01 +03:00
Bundle with full content works
clonebundle: support bundle2 exchange.readbundle() can return 2 different types. We weren't handling the bundle2 case. Handle it. At some point we'll likely want a generic API for applying a bundle from a file handle. For now, create another one-off until we figure out what the unified bundle API should look like (addressing this is a can of worms I don't want to open right now).
2015-10-13 20:41:54 +03:00
$ hg -R server bundle --type gzip-v2 --base null -r tip full.hg
clonebundles: support for seeding clones from pre-generated bundles Cloning can be an expensive operation for servers because the server generates a bundle from existing repository data at request time. For a large repository like mozilla-central, this consumes 4+ minutes of CPU time on the server. It also results in significant network utilization. Multiplied by hundreds or even thousands of clients and the ensuing load can result in difficulties scaling the Mercurial server. Despite generation of bundles being deterministic until the next changeset is added, the generation of bundles to service a clone request is not cached. Each clone thus performs redundant work. This is wasteful. This patch introduces the "clonebundles" extension and related client-side functionality to help alleviate this deficiency. The client-side feature is behind an experimental flag and is not enabled by default. It works as follows: 1) Server operator generates a bundle and makes it available on a server (likely HTTP). 2) Server operator defines the URL of a bundle file in a .hg/clonebundles.manifest file. 3) Client `hg clone`ing sees the server is advertising bundle URLs. 4) Client fetches and applies the advertised bundle. 5) Client performs equivalent of `hg pull` to fetch changes made since the bundle was created. Essentially, the server performs the expensive work of generating a bundle once and all subsequent clones fetch a static file from somewhere. Scaling static file serving is a much more manageable problem than scaling a Python application like Mercurial. Assuming your repository grows less than 1% per day, the end result is 99+% of CPU and network load from clones is eliminated, allowing Mercurial servers to scale more easily. Serving static files also means data can be transferred to clients as fast as they can consume it, rather than as fast as servers can generate it. This makes clones faster. Mozilla has implemented similar functionality of this patch on hg.mozilla.org using a custom extension. We are hosting bundle files in Amazon S3 and CloudFront (a CDN) and have successfully offloaded >1 TB/day in data transfer from hg.mozilla.org, freeing up significant bandwidth and CPU resources. The positive impact has been stellar and I believe it has proved its value to be included in Mercurial core. I feel it is important for the client-side support to be enabled in core by default because it means that clients will get faster, more reliable clones and will enable server operators to reduce load without requiring any client-side configuration changes (assuming clients are up to date, of course). The scope of this feature is narrowly and specifically tailored to cloning, despite "serve pulls from pre-generated bundles" being a valid and useful feature. I would eventually like for Mercurial servers to support transferring *all* repository data via statically hosted files. You could imagine a server that siphons all pushed data to bundle files and instructs clients to apply a stream of bundles to reconstruct all repository data. This feature, while useful and powerful, is significantly more work to implement because it requires the server component have awareness of discovery and a mapping of which changesets are in which files. Full, clone bundles, by contrast, are much simpler. The wire protocol command is named "clonebundles" instead of something more generic like "staticbundles" to leave the door open for a new, more powerful and more generic server-side component with minimal backwards compatibility implications. The name "bundleclone" is used by Mozilla's extension and would cause problems since there are subtle differences in Mozilla's extension. Mozilla's experience with this idea has taught us that some form of "content negotiation" is required. Not all clients will support all bundle formats or even URLs (advanced TLS requirements, etc). To ensure the highest uptake possible, a server needs to advertise multiple versions of bundles and clients need to be able to choose the most appropriate from that list one. The "attributes" in each server-advertised entry facilitate this filtering and sorting. Their use will become apparent in subsequent patches. Initial inspiration and credit for the idea of cloning from static files belongs to Augie Fackler and his "lookaside clone" extension proof of concept.
2015-10-09 21:22:01 +03:00
2 changesets found
clonebundles: filter on bundle specification Not all clients are capable of reading every bundle. Currently, content negotiation to ensure a server sends a client a compatible bundle format is performed at request time. The response bundle is dynamically generated at request time, so this works fine. Clone bundles are statically generated *before* the request. This means that a modern server could produce bundles that a legacy client isn't capable of reading. Without some kind of "type hint" in the clone bundles manifest, a client may attempt to download an incompatible bundle. Furthermore, a client may not realize a bundle is incompatible until it has processed part of the bundle (imagine consuming a 1 GB changegroup bundle2 part only to discover the bundle2 part afterwards is incompatibl). This would waste time and resources. And it isn't very user friendly. Clone bundle manifests thus need to advertise the *exact* format of the hosted bundles so clients may filter out entries that they don't know how to read. This patch introduces that mechanism. We introduce the BUNDLESPEC attribute to declare the "bundle specification" of the entry. Bundle specifications are parsed using exchange.parsebundlespecification, which uses the same strings as the "--type" argument to `hg bundle`. The supported bundle specifications are well defined and backwards compatible. When a client encounters a BUNDLESPEC that is invalid or unsupported, it silently ignores the entry.
2015-10-13 21:45:30 +03:00
Again, we perform an extra check against bundle content changes. If this content
changes, clone bundles produced by new Mercurial versions may not be readable
by old clients.
$ f --size --hexdump full.hg
mdiff: don't emit a diff header for empty trivial deltas An empty trivial delta, coded as (0, 0, 0) makes the delta application do nothing, but still takes 12 bytes, while skipping it altogether works as much, without taking any space at all.
2016-01-12 06:00:07 +03:00
full.hg: size=406
clonebundles: filter on bundle specification Not all clients are capable of reading every bundle. Currently, content negotiation to ensure a server sends a client a compatible bundle format is performed at request time. The response bundle is dynamically generated at request time, so this works fine. Clone bundles are statically generated *before* the request. This means that a modern server could produce bundles that a legacy client isn't capable of reading. Without some kind of "type hint" in the clone bundles manifest, a client may attempt to download an incompatible bundle. Furthermore, a client may not realize a bundle is incompatible until it has processed part of the bundle (imagine consuming a 1 GB changegroup bundle2 part only to discover the bundle2 part afterwards is incompatibl). This would waste time and resources. And it isn't very user friendly. Clone bundle manifests thus need to advertise the *exact* format of the hosted bundles so clients may filter out entries that they don't know how to read. This patch introduces that mechanism. We introduce the BUNDLESPEC attribute to declare the "bundle specification" of the entry. Bundle specifications are parsed using exchange.parsebundlespecification, which uses the same strings as the "--type" argument to `hg bundle`. The supported bundle specifications are well defined and backwards compatible. When a client encounters a BUNDLESPEC that is invalid or unsupported, it silently ignores the entry.
2015-10-13 21:45:30 +03:00
0000: 48 47 32 30 00 00 00 0e 43 6f 6d 70 72 65 73 73 |HG20....Compress|
0010: 69 6f 6e 3d 47 5a 78 9c 63 60 60 90 e5 76 f6 70 |ion=GZx.c``..v.p|
0020: f4 73 77 75 0f f2 0f 0d 60 00 02 46 06 76 a6 b2 |.swu....`..F.v..|
mdiff: don't emit a diff header for empty trivial deltas An empty trivial delta, coded as (0, 0, 0) makes the delta application do nothing, but still takes 12 bytes, while skipping it altogether works as much, without taking any space at all.
2016-01-12 06:00:07 +03:00
0030: d4 a2 e2 cc fc 3c 03 23 06 06 e6 65 40 b1 4d c1 |.....<.#...e@.M.|
clonebundles: filter on bundle specification Not all clients are capable of reading every bundle. Currently, content negotiation to ensure a server sends a client a compatible bundle format is performed at request time. The response bundle is dynamically generated at request time, so this works fine. Clone bundles are statically generated *before* the request. This means that a modern server could produce bundles that a legacy client isn't capable of reading. Without some kind of "type hint" in the clone bundles manifest, a client may attempt to download an incompatible bundle. Furthermore, a client may not realize a bundle is incompatible until it has processed part of the bundle (imagine consuming a 1 GB changegroup bundle2 part only to discover the bundle2 part afterwards is incompatibl). This would waste time and resources. And it isn't very user friendly. Clone bundle manifests thus need to advertise the *exact* format of the hosted bundles so clients may filter out entries that they don't know how to read. This patch introduces that mechanism. We introduce the BUNDLESPEC attribute to declare the "bundle specification" of the entry. Bundle specifications are parsed using exchange.parsebundlespecification, which uses the same strings as the "--type" argument to `hg bundle`. The supported bundle specifications are well defined and backwards compatible. When a client encounters a BUNDLESPEC that is invalid or unsupported, it silently ignores the entry.
2015-10-13 21:45:30 +03:00
0040: 2a 31 09 cf 9a 3a 52 04 b7 fc db f0 95 e5 a4 f4 |*1...:R.........|
0050: 97 17 b2 c9 0c 14 00 02 e6 d9 99 25 1a a7 a4 99 |...........%....|
0060: a4 a4 1a 5b 58 a4 19 27 9b a4 59 a4 1a 59 a4 99 |...[X..'..Y..Y..|
0070: a4 59 26 5a 18 9a 18 59 5a 26 1a 27 27 25 99 a6 |.Y&Z...YZ&.''%..|
0080: 99 1a 70 95 a4 16 97 70 19 28 18 70 a5 e5 e7 73 |..p....p.(.p...s|
0090: 71 25 a6 a4 28 00 19 40 13 0e ac fa df ab ff 7b |q%..(..@.......{|
00a0: 3f fb 92 dc 8b 1f 62 bb 9e b7 d7 d9 87 3d 5a 44 |?.....b......=ZD|
00b0: ac 2f b0 a9 c3 66 1e 54 b9 26 08 a7 1a 1b 1a a7 |./...f.T.&......|
00c0: 25 1b 9a 1b 99 19 9a 5a 18 9b a6 18 19 00 dd 67 |%......Z.......g|
00d0: 61 61 98 06 f4 80 49 4a 8a 65 52 92 41 9a 81 81 |aa....IJ.eR.A...|
00e0: a5 11 17 50 31 30 58 19 cc 80 98 25 29 b1 08 c4 |...P10X....%)...|
00f0: 37 07 79 19 88 d9 41 ee 07 8a 41 cd 5d 98 65 fb |7.y...A...A.].e.|
0100: e5 9e 45 bf 8d 7f 9f c6 97 9f 2b 44 34 67 d9 ec |..E.......+D4g..|
0110: 8e 0f a0 61 a8 eb 82 82 2e c9 c2 20 25 d5 34 c5 |...a....... %.4.|
0120: d0 d8 c2 dc d4 c2 d4 c4 30 d9 34 cd c0 d4 c8 cc |........0.4.....|
0130: 34 31 c5 d0 c4 24 31 c9 32 2d d1 c2 2c c5 30 25 |41...$1.2-..,.0%|
0140: 09 e4 ee 85 8f 85 ff 88 ab 89 36 c7 2a c4 47 34 |..........6.*.G4|
0150: fe f8 ec 7b 73 37 3f c3 24 62 1d 8d 4d 1d 9e 40 |...{s7?.$b..M..@|
mdiff: don't emit a diff header for empty trivial deltas An empty trivial delta, coded as (0, 0, 0) makes the delta application do nothing, but still takes 12 bytes, while skipping it altogether works as much, without taking any space at all.
2016-01-12 06:00:07 +03:00
0160: 06 3b 10 14 36 a4 38 10 04 d8 21 01 9a b1 83 f7 |.;..6.8...!.....|
clonebundles: filter on bundle specification Not all clients are capable of reading every bundle. Currently, content negotiation to ensure a server sends a client a compatible bundle format is performed at request time. The response bundle is dynamically generated at request time, so this works fine. Clone bundles are statically generated *before* the request. This means that a modern server could produce bundles that a legacy client isn't capable of reading. Without some kind of "type hint" in the clone bundles manifest, a client may attempt to download an incompatible bundle. Furthermore, a client may not realize a bundle is incompatible until it has processed part of the bundle (imagine consuming a 1 GB changegroup bundle2 part only to discover the bundle2 part afterwards is incompatibl). This would waste time and resources. And it isn't very user friendly. Clone bundle manifests thus need to advertise the *exact* format of the hosted bundles so clients may filter out entries that they don't know how to read. This patch introduces that mechanism. We introduce the BUNDLESPEC attribute to declare the "bundle specification" of the entry. Bundle specifications are parsed using exchange.parsebundlespecification, which uses the same strings as the "--type" argument to `hg bundle`. The supported bundle specifications are well defined and backwards compatible. When a client encounters a BUNDLESPEC that is invalid or unsupported, it silently ignores the entry.
2015-10-13 21:45:30 +03:00
0170: e9 45 8b d2 56 c7 a3 1f 82 52 d7 8a 78 ed fc d5 |.E..V....R..x...|
mdiff: don't emit a diff header for empty trivial deltas An empty trivial delta, coded as (0, 0, 0) makes the delta application do nothing, but still takes 12 bytes, while skipping it altogether works as much, without taking any space at all.
2016-01-12 06:00:07 +03:00
0180: 76 f1 36 25 81 89 c7 ad ec 90 34 48 75 2b 89 49 |v.6%......4Hu+.I|
0190: bf 00 d6 97 f0 8d |......|
clonebundles: filter on bundle specification Not all clients are capable of reading every bundle. Currently, content negotiation to ensure a server sends a client a compatible bundle format is performed at request time. The response bundle is dynamically generated at request time, so this works fine. Clone bundles are statically generated *before* the request. This means that a modern server could produce bundles that a legacy client isn't capable of reading. Without some kind of "type hint" in the clone bundles manifest, a client may attempt to download an incompatible bundle. Furthermore, a client may not realize a bundle is incompatible until it has processed part of the bundle (imagine consuming a 1 GB changegroup bundle2 part only to discover the bundle2 part afterwards is incompatibl). This would waste time and resources. And it isn't very user friendly. Clone bundle manifests thus need to advertise the *exact* format of the hosted bundles so clients may filter out entries that they don't know how to read. This patch introduces that mechanism. We introduce the BUNDLESPEC attribute to declare the "bundle specification" of the entry. Bundle specifications are parsed using exchange.parsebundlespecification, which uses the same strings as the "--type" argument to `hg bundle`. The supported bundle specifications are well defined and backwards compatible. When a client encounters a BUNDLESPEC that is invalid or unsupported, it silently ignores the entry.
2015-10-13 21:45:30 +03:00
clonebundles: support for seeding clones from pre-generated bundles Cloning can be an expensive operation for servers because the server generates a bundle from existing repository data at request time. For a large repository like mozilla-central, this consumes 4+ minutes of CPU time on the server. It also results in significant network utilization. Multiplied by hundreds or even thousands of clients and the ensuing load can result in difficulties scaling the Mercurial server. Despite generation of bundles being deterministic until the next changeset is added, the generation of bundles to service a clone request is not cached. Each clone thus performs redundant work. This is wasteful. This patch introduces the "clonebundles" extension and related client-side functionality to help alleviate this deficiency. The client-side feature is behind an experimental flag and is not enabled by default. It works as follows: 1) Server operator generates a bundle and makes it available on a server (likely HTTP). 2) Server operator defines the URL of a bundle file in a .hg/clonebundles.manifest file. 3) Client `hg clone`ing sees the server is advertising bundle URLs. 4) Client fetches and applies the advertised bundle. 5) Client performs equivalent of `hg pull` to fetch changes made since the bundle was created. Essentially, the server performs the expensive work of generating a bundle once and all subsequent clones fetch a static file from somewhere. Scaling static file serving is a much more manageable problem than scaling a Python application like Mercurial. Assuming your repository grows less than 1% per day, the end result is 99+% of CPU and network load from clones is eliminated, allowing Mercurial servers to scale more easily. Serving static files also means data can be transferred to clients as fast as they can consume it, rather than as fast as servers can generate it. This makes clones faster. Mozilla has implemented similar functionality of this patch on hg.mozilla.org using a custom extension. We are hosting bundle files in Amazon S3 and CloudFront (a CDN) and have successfully offloaded >1 TB/day in data transfer from hg.mozilla.org, freeing up significant bandwidth and CPU resources. The positive impact has been stellar and I believe it has proved its value to be included in Mercurial core. I feel it is important for the client-side support to be enabled in core by default because it means that clients will get faster, more reliable clones and will enable server operators to reduce load without requiring any client-side configuration changes (assuming clients are up to date, of course). The scope of this feature is narrowly and specifically tailored to cloning, despite "serve pulls from pre-generated bundles" being a valid and useful feature. I would eventually like for Mercurial servers to support transferring *all* repository data via statically hosted files. You could imagine a server that siphons all pushed data to bundle files and instructs clients to apply a stream of bundles to reconstruct all repository data. This feature, while useful and powerful, is significantly more work to implement because it requires the server component have awareness of discovery and a mapping of which changesets are in which files. Full, clone bundles, by contrast, are much simpler. The wire protocol command is named "clonebundles" instead of something more generic like "staticbundles" to leave the door open for a new, more powerful and more generic server-side component with minimal backwards compatibility implications. The name "bundleclone" is used by Mozilla's extension and would cause problems since there are subtle differences in Mozilla's extension. Mozilla's experience with this idea has taught us that some form of "content negotiation" is required. Not all clients will support all bundle formats or even URLs (advanced TLS requirements, etc). To ensure the highest uptake possible, a server needs to advertise multiple versions of bundles and clients need to be able to choose the most appropriate from that list one. The "attributes" in each server-advertised entry facilitate this filtering and sorting. Their use will become apparent in subsequent patches. Initial inspiration and credit for the idea of cloning from static files belongs to Augie Fackler and his "lookaside clone" extension proof of concept.
2015-10-09 21:22:01 +03:00
$ echo "http://localhost:$HGPORT1/full.hg" > server/.hg/clonebundles.manifest
$ hg clone -U http://localhost:$HGPORT full-bundle
applying clone bundle from http://localhost:$HGPORT1/full.hg
adding changesets
adding manifests
adding file changes
added 2 changesets with 2 changes to 2 files
finished applying clone bundle
searching for changes
no changes found
clonebundles: filter on bundle specification Not all clients are capable of reading every bundle. Currently, content negotiation to ensure a server sends a client a compatible bundle format is performed at request time. The response bundle is dynamically generated at request time, so this works fine. Clone bundles are statically generated *before* the request. This means that a modern server could produce bundles that a legacy client isn't capable of reading. Without some kind of "type hint" in the clone bundles manifest, a client may attempt to download an incompatible bundle. Furthermore, a client may not realize a bundle is incompatible until it has processed part of the bundle (imagine consuming a 1 GB changegroup bundle2 part only to discover the bundle2 part afterwards is incompatibl). This would waste time and resources. And it isn't very user friendly. Clone bundle manifests thus need to advertise the *exact* format of the hosted bundles so clients may filter out entries that they don't know how to read. This patch introduces that mechanism. We introduce the BUNDLESPEC attribute to declare the "bundle specification" of the entry. Bundle specifications are parsed using exchange.parsebundlespecification, which uses the same strings as the "--type" argument to `hg bundle`. The supported bundle specifications are well defined and backwards compatible. When a client encounters a BUNDLESPEC that is invalid or unsupported, it silently ignores the entry.
2015-10-13 21:45:30 +03:00
wireproto: move clonebundles command from extension (issue4931) The SSH peer class accesses wireproto.commands[cmd] as part of encoding command arguments. Previously, the wire protocol command was defined in the clonebundles extension. If the client didn't have this extension enabled (which it likely doesn't since it is meant as a server-side extension), then clients attempting to clone via ssh:// would get a crash due to a KeyError accessing wireproto.commands['clonebundles'] when cloning from a server that is advertising clone bundles. Moving the definition of the wire protocol command to wireproto.py makes this problem go away. A side effect of this code move is servers will always respond to "clonebundles" wire protocol command requests. This should be fine: the server will return an empty response unless a clone bundles manifest file is present and clients shouldn't call the command unless the server is advertising the capability, which only happens if the clonebundles extension is enabled and the manifest file exists.
2015-11-03 23:31:33 +03:00
Feature works over SSH
$ hg clone -U -e "python \"$TESTDIR/dummyssh\"" ssh://user@dummy/server ssh-full-clone
applying clone bundle from http://localhost:$HGPORT1/full.hg
adding changesets
adding manifests
adding file changes
added 2 changesets with 2 changes to 2 files
finished applying clone bundle
searching for changes
no changes found
clonebundles: filter on bundle specification Not all clients are capable of reading every bundle. Currently, content negotiation to ensure a server sends a client a compatible bundle format is performed at request time. The response bundle is dynamically generated at request time, so this works fine. Clone bundles are statically generated *before* the request. This means that a modern server could produce bundles that a legacy client isn't capable of reading. Without some kind of "type hint" in the clone bundles manifest, a client may attempt to download an incompatible bundle. Furthermore, a client may not realize a bundle is incompatible until it has processed part of the bundle (imagine consuming a 1 GB changegroup bundle2 part only to discover the bundle2 part afterwards is incompatibl). This would waste time and resources. And it isn't very user friendly. Clone bundle manifests thus need to advertise the *exact* format of the hosted bundles so clients may filter out entries that they don't know how to read. This patch introduces that mechanism. We introduce the BUNDLESPEC attribute to declare the "bundle specification" of the entry. Bundle specifications are parsed using exchange.parsebundlespecification, which uses the same strings as the "--type" argument to `hg bundle`. The supported bundle specifications are well defined and backwards compatible. When a client encounters a BUNDLESPEC that is invalid or unsupported, it silently ignores the entry.
2015-10-13 21:45:30 +03:00
Entry with unknown BUNDLESPEC is filtered and not used
$ cat > server/.hg/clonebundles.manifest << EOF
> http://bad.entry1 BUNDLESPEC=UNKNOWN
> http://bad.entry2 BUNDLESPEC=xz-v1
> http://bad.entry3 BUNDLESPEC=none-v100
> http://localhost:$HGPORT1/full.hg BUNDLESPEC=gzip-v2
> EOF
$ hg clone -U http://localhost:$HGPORT filter-unknown-type
applying clone bundle from http://localhost:$HGPORT1/full.hg
adding changesets
adding manifests
adding file changes
added 2 changesets with 2 changes to 2 files
finished applying clone bundle
searching for changes
no changes found
Automatic fallback when all entries are filtered
$ cat > server/.hg/clonebundles.manifest << EOF
> http://bad.entry BUNDLESPEC=UNKNOWN
> EOF
$ hg clone -U http://localhost:$HGPORT filter-all
no compatible clone bundles available on server; falling back to regular clone
(you may want to report this to the server operator)
requesting all changes
adding changesets
adding manifests
adding file changes
added 2 changesets with 2 changes to 2 files
clonebundles: filter on SNI requirement Server Name Indication (SNI) is commonly used in CDNs and other hosted environments. Unfortunately, Python <2.7.9 does not support SNI and when these older Python versions attempt to negotiate TLS to an SNI server, they raise an opaque error like "_ssl.c:507: error:14094410:SSL routines:SSL3_READ_BYTES:sslv3 alert handshake failure." We introduce a manifest attribute to denote the URL requires SNI and have clients without SNI support filter these entries.
2015-10-13 20:59:41 +03:00
URLs requiring SNI are filtered in Python <2.7.9
$ cp full.hg sni.hg
$ cat > server/.hg/clonebundles.manifest << EOF
> http://localhost:$HGPORT1/sni.hg REQUIRESNI=true
> http://localhost:$HGPORT1/full.hg
> EOF
#if sslcontext
Python 2.7.9+ support SNI
$ hg clone -U http://localhost:$HGPORT sni-supported
applying clone bundle from http://localhost:$HGPORT1/sni.hg
adding changesets
adding manifests
adding file changes
added 2 changesets with 2 changes to 2 files
finished applying clone bundle
searching for changes
no changes found
#else
Python <2.7.9 will filter SNI URLs
$ hg clone -U http://localhost:$HGPORT sni-unsupported
applying clone bundle from http://localhost:$HGPORT1/full.hg
adding changesets
adding manifests
adding file changes
added 2 changesets with 2 changes to 2 files
finished applying clone bundle
searching for changes
no changes found
#endif
exchange: support sorting URLs by client-side preferences Not all bundles are appropriate for all clients. For example, someone with a slow Internet connection may want to prefer bz2 bundles over gzip bundles because they are smaller and don't take as long to transfer. This is information that a server cannot know on its own. So, we invent a mechanism for "preferring" server-advertised URLs based on their attributes. We could invent a negotiation between client and server where the client sends its preferences and the sorting/filtering is done server-side. However, this feels complex. We can avoid complicating the wire protocol and exposing ourselves to backwards compatible concerns by performing the sorting locally. This patch defines a new config option for expressing preferred attributes in server-advertised bundles. At Mozilla, we leverage this feature so clients in fast data centers prefer uncompressed bundles. (We advertise gzip bundles first because that is a reasonable default.) I consider this an advanced feature. I'm on the fence as to whether it should be documented in `hg help config`.
2015-10-13 22:30:39 +03:00
exchange: support streaming clone bundles in clone bundles Now that we have support for detecting compatible stream clone bundles in bundle specifications, we can safely add support for applying stream clone bundles to the clone bundles feature.
2015-10-17 21:37:08 +03:00
Stream clone bundles are supported
$ hg -R server debugcreatestreamclonebundle packed.hg
writing 613 bytes for 4 files
test: use generaldelta in 'test-clonebundle.t' Using generaldelta has effects on the generated bundle.
2015-11-07 21:48:42 +03:00
bundle requirements: generaldelta, revlogv1
exchange: support streaming clone bundles in clone bundles Now that we have support for detecting compatible stream clone bundles in bundle specifications, we can safely add support for applying stream clone bundles to the clone bundles feature.
2015-10-17 21:37:08 +03:00
No bundle spec should work
$ cat > server/.hg/clonebundles.manifest << EOF
> http://localhost:$HGPORT1/packed.hg
> EOF
$ hg clone -U http://localhost:$HGPORT stream-clone-no-spec
applying clone bundle from http://localhost:$HGPORT1/packed.hg
4 files to transfer, 613 bytes of data
transferred 613 bytes in *.* seconds (*) (glob)
finished applying clone bundle
searching for changes
no changes found
Bundle spec without parameters should work
$ cat > server/.hg/clonebundles.manifest << EOF
> http://localhost:$HGPORT1/packed.hg BUNDLESPEC=none-packed1
> EOF
$ hg clone -U http://localhost:$HGPORT stream-clone-vanilla-spec
applying clone bundle from http://localhost:$HGPORT1/packed.hg
4 files to transfer, 613 bytes of data
transferred 613 bytes in *.* seconds (*) (glob)
finished applying clone bundle
searching for changes
no changes found
Bundle spec with format requirements should work
$ cat > server/.hg/clonebundles.manifest << EOF
> http://localhost:$HGPORT1/packed.hg BUNDLESPEC=none-packed1;requirements%3Drevlogv1
> EOF
$ hg clone -U http://localhost:$HGPORT stream-clone-supported-requirements
applying clone bundle from http://localhost:$HGPORT1/packed.hg
4 files to transfer, 613 bytes of data
transferred 613 bytes in *.* seconds (*) (glob)
finished applying clone bundle
searching for changes
no changes found
Stream bundle spec with unknown requirements should be filtered out
$ cat > server/.hg/clonebundles.manifest << EOF
> http://localhost:$HGPORT1/packed.hg BUNDLESPEC=none-packed1;requirements%3Drevlogv42
> EOF
$ hg clone -U http://localhost:$HGPORT stream-clone-unsupported-requirements
no compatible clone bundles available on server; falling back to regular clone
(you may want to report this to the server operator)
requesting all changes
adding changesets
adding manifests
adding file changes
added 2 changesets with 2 changes to 2 files
exchange: support sorting URLs by client-side preferences Not all bundles are appropriate for all clients. For example, someone with a slow Internet connection may want to prefer bz2 bundles over gzip bundles because they are smaller and don't take as long to transfer. This is information that a server cannot know on its own. So, we invent a mechanism for "preferring" server-advertised URLs based on their attributes. We could invent a negotiation between client and server where the client sends its preferences and the sorting/filtering is done server-side. However, this feels complex. We can avoid complicating the wire protocol and exposing ourselves to backwards compatible concerns by performing the sorting locally. This patch defines a new config option for expressing preferred attributes in server-advertised bundles. At Mozilla, we leverage this feature so clients in fast data centers prefer uncompressed bundles. (We advertise gzip bundles first because that is a reasonable default.) I consider this an advanced feature. I'm on the fence as to whether it should be documented in `hg help config`.
2015-10-13 22:30:39 +03:00
Set up manifest for testing preferences
(Remember, the TYPE does not have to match reality - the URL is
important)
$ cp full.hg gz-a.hg
$ cp full.hg gz-b.hg
$ cp full.hg bz2-a.hg
$ cp full.hg bz2-b.hg
$ cat > server/.hg/clonebundles.manifest << EOF
> http://localhost:$HGPORT1/gz-a.hg BUNDLESPEC=gzip-v2 extra=a
> http://localhost:$HGPORT1/bz2-a.hg BUNDLESPEC=bzip2-v2 extra=a
> http://localhost:$HGPORT1/gz-b.hg BUNDLESPEC=gzip-v2 extra=b
> http://localhost:$HGPORT1/bz2-b.hg BUNDLESPEC=bzip2-v2 extra=b
> EOF
Preferring an undefined attribute will take first entry
exchange: make clonebundleprefers non-experimental In preparation for making the feature enabled by default.
2016-01-08 21:57:01 +03:00
$ hg --config ui.clonebundleprefers=foo=bar clone -U http://localhost:$HGPORT prefer-foo
exchange: support sorting URLs by client-side preferences Not all bundles are appropriate for all clients. For example, someone with a slow Internet connection may want to prefer bz2 bundles over gzip bundles because they are smaller and don't take as long to transfer. This is information that a server cannot know on its own. So, we invent a mechanism for "preferring" server-advertised URLs based on their attributes. We could invent a negotiation between client and server where the client sends its preferences and the sorting/filtering is done server-side. However, this feels complex. We can avoid complicating the wire protocol and exposing ourselves to backwards compatible concerns by performing the sorting locally. This patch defines a new config option for expressing preferred attributes in server-advertised bundles. At Mozilla, we leverage this feature so clients in fast data centers prefer uncompressed bundles. (We advertise gzip bundles first because that is a reasonable default.) I consider this an advanced feature. I'm on the fence as to whether it should be documented in `hg help config`.
2015-10-13 22:30:39 +03:00
applying clone bundle from http://localhost:$HGPORT1/gz-a.hg
adding changesets
adding manifests
adding file changes
added 2 changesets with 2 changes to 2 files
finished applying clone bundle
searching for changes
no changes found
Preferring bz2 type will download first entry of that type
exchange: make clonebundleprefers non-experimental In preparation for making the feature enabled by default.
2016-01-08 21:57:01 +03:00
$ hg --config ui.clonebundleprefers=COMPRESSION=bzip2 clone -U http://localhost:$HGPORT prefer-bz
exchange: support sorting URLs by client-side preferences Not all bundles are appropriate for all clients. For example, someone with a slow Internet connection may want to prefer bz2 bundles over gzip bundles because they are smaller and don't take as long to transfer. This is information that a server cannot know on its own. So, we invent a mechanism for "preferring" server-advertised URLs based on their attributes. We could invent a negotiation between client and server where the client sends its preferences and the sorting/filtering is done server-side. However, this feels complex. We can avoid complicating the wire protocol and exposing ourselves to backwards compatible concerns by performing the sorting locally. This patch defines a new config option for expressing preferred attributes in server-advertised bundles. At Mozilla, we leverage this feature so clients in fast data centers prefer uncompressed bundles. (We advertise gzip bundles first because that is a reasonable default.) I consider this an advanced feature. I'm on the fence as to whether it should be documented in `hg help config`.
2015-10-13 22:30:39 +03:00
applying clone bundle from http://localhost:$HGPORT1/bz2-a.hg
adding changesets
adding manifests
adding file changes
added 2 changesets with 2 changes to 2 files
finished applying clone bundle
searching for changes
no changes found
Preferring multiple values of an option works
exchange: make clonebundleprefers non-experimental In preparation for making the feature enabled by default.
2016-01-08 21:57:01 +03:00
$ hg --config ui.clonebundleprefers=COMPRESSION=unknown,COMPRESSION=bzip2 clone -U http://localhost:$HGPORT prefer-multiple-bz
exchange: support sorting URLs by client-side preferences Not all bundles are appropriate for all clients. For example, someone with a slow Internet connection may want to prefer bz2 bundles over gzip bundles because they are smaller and don't take as long to transfer. This is information that a server cannot know on its own. So, we invent a mechanism for "preferring" server-advertised URLs based on their attributes. We could invent a negotiation between client and server where the client sends its preferences and the sorting/filtering is done server-side. However, this feels complex. We can avoid complicating the wire protocol and exposing ourselves to backwards compatible concerns by performing the sorting locally. This patch defines a new config option for expressing preferred attributes in server-advertised bundles. At Mozilla, we leverage this feature so clients in fast data centers prefer uncompressed bundles. (We advertise gzip bundles first because that is a reasonable default.) I consider this an advanced feature. I'm on the fence as to whether it should be documented in `hg help config`.
2015-10-13 22:30:39 +03:00
applying clone bundle from http://localhost:$HGPORT1/bz2-a.hg
adding changesets
adding manifests
adding file changes
added 2 changesets with 2 changes to 2 files
finished applying clone bundle
searching for changes
no changes found
Sorting multiple values should get us back to original first entry
exchange: make clonebundleprefers non-experimental In preparation for making the feature enabled by default.
2016-01-08 21:57:01 +03:00
$ hg --config ui.clonebundleprefers=BUNDLESPEC=unknown,BUNDLESPEC=gzip-v2,BUNDLESPEC=bzip2-v2 clone -U http://localhost:$HGPORT prefer-multiple-gz
exchange: support sorting URLs by client-side preferences Not all bundles are appropriate for all clients. For example, someone with a slow Internet connection may want to prefer bz2 bundles over gzip bundles because they are smaller and don't take as long to transfer. This is information that a server cannot know on its own. So, we invent a mechanism for "preferring" server-advertised URLs based on their attributes. We could invent a negotiation between client and server where the client sends its preferences and the sorting/filtering is done server-side. However, this feels complex. We can avoid complicating the wire protocol and exposing ourselves to backwards compatible concerns by performing the sorting locally. This patch defines a new config option for expressing preferred attributes in server-advertised bundles. At Mozilla, we leverage this feature so clients in fast data centers prefer uncompressed bundles. (We advertise gzip bundles first because that is a reasonable default.) I consider this an advanced feature. I'm on the fence as to whether it should be documented in `hg help config`.
2015-10-13 22:30:39 +03:00
applying clone bundle from http://localhost:$HGPORT1/gz-a.hg
adding changesets
adding manifests
adding file changes
added 2 changesets with 2 changes to 2 files
finished applying clone bundle
searching for changes
no changes found
Preferring multiple attributes has correct order
exchange: make clonebundleprefers non-experimental In preparation for making the feature enabled by default.
2016-01-08 21:57:01 +03:00
$ hg --config ui.clonebundleprefers=extra=b,BUNDLESPEC=bzip2-v2 clone -U http://localhost:$HGPORT prefer-separate-attributes
exchange: support sorting URLs by client-side preferences Not all bundles are appropriate for all clients. For example, someone with a slow Internet connection may want to prefer bz2 bundles over gzip bundles because they are smaller and don't take as long to transfer. This is information that a server cannot know on its own. So, we invent a mechanism for "preferring" server-advertised URLs based on their attributes. We could invent a negotiation between client and server where the client sends its preferences and the sorting/filtering is done server-side. However, this feels complex. We can avoid complicating the wire protocol and exposing ourselves to backwards compatible concerns by performing the sorting locally. This patch defines a new config option for expressing preferred attributes in server-advertised bundles. At Mozilla, we leverage this feature so clients in fast data centers prefer uncompressed bundles. (We advertise gzip bundles first because that is a reasonable default.) I consider this an advanced feature. I'm on the fence as to whether it should be documented in `hg help config`.
2015-10-13 22:30:39 +03:00
applying clone bundle from http://localhost:$HGPORT1/bz2-b.hg
adding changesets
adding manifests
adding file changes
added 2 changesets with 2 changes to 2 files
finished applying clone bundle
searching for changes
no changes found
Test where attribute is missing from some entries
$ cat > server/.hg/clonebundles.manifest << EOF
> http://localhost:$HGPORT1/gz-a.hg BUNDLESPEC=gzip-v2
> http://localhost:$HGPORT1/bz2-a.hg BUNDLESPEC=bzip2-v2
> http://localhost:$HGPORT1/gz-b.hg BUNDLESPEC=gzip-v2 extra=b
> http://localhost:$HGPORT1/bz2-b.hg BUNDLESPEC=bzip2-v2 extra=b
> EOF
exchange: make clonebundleprefers non-experimental In preparation for making the feature enabled by default.
2016-01-08 21:57:01 +03:00
$ hg --config ui.clonebundleprefers=extra=b clone -U http://localhost:$HGPORT prefer-partially-defined-attribute
exchange: support sorting URLs by client-side preferences Not all bundles are appropriate for all clients. For example, someone with a slow Internet connection may want to prefer bz2 bundles over gzip bundles because they are smaller and don't take as long to transfer. This is information that a server cannot know on its own. So, we invent a mechanism for "preferring" server-advertised URLs based on their attributes. We could invent a negotiation between client and server where the client sends its preferences and the sorting/filtering is done server-side. However, this feels complex. We can avoid complicating the wire protocol and exposing ourselves to backwards compatible concerns by performing the sorting locally. This patch defines a new config option for expressing preferred attributes in server-advertised bundles. At Mozilla, we leverage this feature so clients in fast data centers prefer uncompressed bundles. (We advertise gzip bundles first because that is a reasonable default.) I consider this an advanced feature. I'm on the fence as to whether it should be documented in `hg help config`.
2015-10-13 22:30:39 +03:00
applying clone bundle from http://localhost:$HGPORT1/gz-b.hg
adding changesets
adding manifests
adding file changes
added 2 changesets with 2 changes to 2 files
finished applying clone bundle
searching for changes
no changes found
Reference in New Issue Copy Permalink