sapling/eden/integration/stats_test.py
Saurabh Singh cfe084d02f telemetry: switch to using quantile stats instead of timeseries
Summary:
Timeseries is memory intensive and not really required in the current context
it is being used.

Reviewed By: chadaustin

Differential Revision: D26315632

fbshipit-source-id: ee51c3ad8bef6fce152aa787c8c4602f0b499f92
2021-02-14 16:37:08 -08:00

256 lines
9.1 KiB
Python

#!/usr/bin/env python3
# Copyright (c) Facebook, Inc. and its affiliates.
#
# This software may be used and distributed according to the terms of the
# GNU General Public License version 2.
import logging
import os
import time
import typing
from pathlib import Path, PurePath
from facebook.eden.ttypes import JournalInfo
from .lib import testcase
from .lib.hgrepo import HgRepository
Counters = typing.Mapping[str, float]
logger = logging.getLogger(__name__)
class FUSEStatsTest(testcase.EdenRepoTest):
def test_reading_committed_file_bumps_read_counter(self) -> None:
counters_before = self.get_counters()
path = Path(self.mount) / "file"
path.read_bytes()
self.poll_until_counter_condition(
lambda counters_after: self.assertGreater(
counters_after.get("fuse.read_us.count", 0),
counters_before.get("fuse.read_us.count", 0),
f"Reading {path} should increment fuse.read_us.count",
)
)
def test_writing_untracked_file_bumps_write_counter(self) -> None:
counters_before = self.get_counters()
path = Path(self.mount) / "new_file"
path.write_bytes(b"hello")
self.poll_until_counter_condition(
lambda counters_after: self.assertGreater(
counters_after.get("fuse.write_us.count", 0),
counters_before.get("fuse.write_us.count", 0),
f"Writing to {path} should increment fuse.write_us.count",
)
)
def test_summary_counters_available(self) -> None:
mountName = PurePath(self.mount).name
counter_names_to_check = [
f"fuse.{mountName}.live_requests.count",
f"fuse.{mountName}.live_requests.max_duration_us",
f"fuse.{mountName}.pending_requests.count",
]
counters = self.get_counters()
for counter_name in counter_names_to_check:
self.assertIn(counter_name, counters, f"{counter_name} should be available")
def create_repo(self, name: str) -> HgRepository:
return self.create_hg_repo(name)
def populate_repo(self) -> None:
self.repo.write_file("file", "hello world!\n")
self.repo.commit("Initial commit.")
def poll_until_counter_condition(
self, assertion_condition: typing.Callable[[Counters], None]
) -> None:
timeout_seconds = 2.0
poll_interval_seconds = 0.1
deadline = time.monotonic() + timeout_seconds
while True:
counters = self.get_counters()
try:
assertion_condition(counters)
break
except AssertionError as e:
if time.monotonic() >= deadline:
raise
logger.info(
f"Assertion failed. Waiting {poll_interval_seconds} "
f"seconds before trying again. {e}"
)
time.sleep(poll_interval_seconds)
continue
class ObjectStoreStatsTest(testcase.EdenRepoTest):
def create_repo(self, name: str) -> HgRepository:
return self.create_hg_repo(name)
def populate_repo(self) -> None:
self.repo.write_file("foo.txt", "foo\n")
self.repo.commit("Initial commit.")
def test_get_blob(self) -> None:
TEMPLATE = "object_store.get_blob.{}_store.rate"
LOCAL = TEMPLATE.format("local")
BACKING = TEMPLATE.format("backing")
counters = self.get_counters()
self.assertEqual(counters.get(LOCAL, 0) + counters.get(BACKING, 0), 0)
foo = Path(self.mount) / "foo.txt"
foo.read_bytes()
counters = self.get_counters()
self.assertEqual(counters.get(LOCAL, 0) + counters.get(BACKING, 0), 2)
class HgBackingStoreStatsTest(testcase.EdenRepoTest):
def test_reading_file_gets_file_from_hg(self) -> None:
counters_before = self.get_counters()
path = Path(self.mount) / "dir" / "subdir" / "file"
path.read_bytes()
counters_after = self.get_counters()
self.assertEqual(
counters_after.get("store.hg.get_blob.count", 0),
counters_before.get("store.hg.get_blob.count", 0) + 1,
f"Reading {path} should increment store.hg.get_blob.count",
)
self.assertEqual(
counters_after.get("store.hg.import_blob.count", 0),
counters_before.get("store.hg.import_blob.count", 0) + 1,
f"Reading {path} should increment store.hg.import_blob.count",
)
def test_pending_import_counters_available(self) -> None:
counters = self.get_counters()
counter_names_to_check = [
"store.hg.pending_import.blob.count",
"store.hg.pending_import.tree.count",
"store.hg.pending_import.prefetch.count",
"store.hg.pending_import.count",
"store.hg.pending_import.blob.max_duration_us",
"store.hg.pending_import.tree.max_duration_us",
"store.hg.pending_import.prefetch.max_duration_us",
"store.hg.pending_import.max_duration_us",
"store.hg.live_import.blob.count",
"store.hg.live_import.tree.count",
"store.hg.live_import.prefetch.count",
"store.hg.live_import.count",
"store.hg.live_import.blob.max_duration_us",
"store.hg.live_import.tree.max_duration_us",
"store.hg.live_import.prefetch.max_duration_us",
"store.hg.live_import.max_duration_us",
]
for counter_name in counter_names_to_check:
self.assertIn(counter_name, counters, f"{counter_name} should be available")
def create_repo(self, name: str) -> HgRepository:
return self.create_hg_repo(name)
def populate_repo(self) -> None:
# This file evades EdenFS' automatic prefetching by being two levels
# inside the root.
self.repo.write_file("dir/subdir/file", "hello world!\n")
self.repo.commit("Initial commit.")
class HgImporterStatsTest(testcase.EdenRepoTest):
def test_reading_file_imports_blob(self) -> None:
counters_before = self.get_counters()
path = Path(self.mount) / "dir" / "subdir" / "file"
path.read_bytes()
counters_after = self.get_counters()
self.assertEqual(
counters_after.get("hg_importer.cat_file.count", 0),
counters_before.get("hg_importer.cat_file.count", 0) + 1,
f"Reading {path} should increment hg_importer.cat_file.count",
)
def create_repo(self, name: str) -> HgRepository:
return self.create_hg_repo(name)
def populate_repo(self) -> None:
# This file evades EdenFS' automatic prefetching by being two levels
# inside the root.
self.repo.write_file("dir/subdir/file", "hello world!\n")
self.repo.commit("Initial commit.")
@testcase.eden_repo_test
class JournalInfoTest(testcase.EdenRepoTest):
def test_journal_info(self) -> None:
journal = self.journal_stats()
old_mem = journal.memoryUsage
old_data_counts = journal.entryCount
path = Path(self.mount) / "new_file"
path.write_bytes(b"hello")
journal = self.journal_stats()
new_mem = journal.memoryUsage
new_data_counts = journal.entryCount
self.assertLess(
old_data_counts,
new_data_counts,
"Changing the repo should cause entry count to increase",
)
self.assertLess(
old_mem, new_mem, "Changing the repo should cause memory usage to increase"
)
def journal_stats(self) -> JournalInfo:
with self.get_thrift_client() as thrift_client:
stats = thrift_client.getStatInfo()
journal_key = self.mount.encode()
journal = stats.mountPointJournalInfo[journal_key]
self.assertIsNotNone(journal, "Journal does not exist")
return journal
def populate_repo(self) -> None:
self.repo.write_file("file", "hello world!\n")
self.repo.commit("Initial commit.")
@testcase.eden_repo_test
class CountersTest(testcase.EdenRepoTest):
"""Test counters are registered/unregistered correctly."""
def populate_repo(self) -> None:
self.repo.write_file("hello", "hola\n")
self.repo.commit("Initial commit.")
# We get rid of the thrift counters since they sporadically appear and can
# cause this test to fail (since they can appear between counters and counters2)
@staticmethod
def get_nonthrift_set(s):
# and memory_vm_rss_bytes is reported sporadically in the background
return {
item
for item in s
if not item.startswith("thrift")
and not item.startswith("memory_vm_rss_bytes")
}
def test_mount_unmount_counters(self) -> None:
self.eden.unmount(self.mount_path)
counters = self.get_nonthrift_set(self.get_counters().keys())
mount2 = os.path.join(self.mounts_dir, "mount2")
self.eden.clone(self.repo.path, mount2)
self.eden.unmount(Path(mount2))
counters2 = self.get_nonthrift_set(self.get_counters().keys())
self.assertEqual(counters, counters2)