urbit/shrub
1
1
Fork 0
mirror of https://github.com/urbit/shrub.git synced 2025-01-01 17:16:47 +03:00
Code Issues Packages Projects Releases Wiki Activity

Event log replay succeeds.

Browse Source
This commit is contained in:
Elliot Glaysher 2019-04-23 14:15:11 -07:00
parent 281e352c14
commit 8b4f1d8e93
2 changed files with 26 additions and 20 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
pkg/urbit/vere/lmdb.c
View File

@ -254,9 +254,6 @@ void u3m_lmdb_write_events(MDB_env* environment,
c3_y* data_u = c3_calloc(siz_w);
u3r_bytes(0, siz_w, data_u, event_u->mat);
fprintf(stderr, "writing event id %" PRIu64 " to db with len %u\r\n",
event_u->evt_d, siz_w);
// Structure to pass to the worker thread.
struct _write_request_data* data = c3_malloc(sizeof(struct _write_request_data));
data->environment = environment;

43
pkg/urbit/vere/pier.c
View File

@ -350,9 +350,6 @@ _pier_db_on_commit_loaded(u3_pier* pir_u,
return c3n;
}
fprintf(stderr, "replaying event %" PRIu64 "\r\n", id);
// Need to grab references to the nouns above.
u3_writ* wit_u = c3_calloc(sizeof(u3_writ));
wit_u->pir_u = pir_u;
@ -360,8 +357,6 @@ _pier_db_on_commit_loaded(u3_pier* pir_u,
wit_u->job = u3k(job);
wit_u->mat = u3k(mat);
// OK, so my queuing logic isn't correct, since I'm never setting the exit
// Insert at queue front since we're loading events in order
if ( !pir_u->ent_u ) {
c3_assert(!pir_u->ext_u);
@ -392,9 +387,6 @@ _pier_db_load_commits(u3_pier* pir_u,
c3_d lav_d,
c3_d len_d)
{
u3l_log("pier: replaying events from %" PRIu64 " (count %" PRIu64 ")\r\n",
lav_d, len_d);
if (lav_d == 1) {
// We are restarting from event 1. That means we need to set the ship from
// the log identity information.
@ -472,12 +464,6 @@ _pier_disk_load_commit(u3_pier* pir_u,
evt_d = u3r_chub(0, evt);
u3z(ovo);
fprintf(stderr, "Loaded %" PRIu64 " with size %llu\r\n", evt_d, len_d);
/* u3m_p("mat", mat); */
/* fprintf(stderr, "Loaded %" PRIu64 " with size %zu\r\n", evt_d, len_d); */
// confirm event order
//
if ( (0 != old_d) &&
@ -877,11 +863,33 @@ _pier_work_save(u3_pier* pir_u)
u3_disk* log_u = pir_u->log_u;
u3_save* sav_u = pir_u->sav_u;
c3_assert( god_u->dun_d == sav_u->req_d );
// This is wrong? We can have requested that the worker should snapshot
// between when the worker was given a compute request and when the worker
// returns a compute response. (See the c3_max() call in u3_pier_snap().) I
// don't know what the semantics are supposed to be here.
//
// Let's say we're in replay of the event log from nothing. We might have
// u3_pier_snap() called from a timer. I've seen crashes when:
//
// god_u->dun_d: 24, sav_u->req_d: 25, log_u->com_d: 31
//
// com_d is greater than req_d is greater than dun_d because we're waiting
// for the worker to complete replay of already committed events. And this
// seems OK because we can totally request that the worker produce a snapshot
// of event 25 even though we haven't marked it as done because we know that
// it committed previously?
//
// No! This breaks because it ignores the req_d and instead tells the worker
// to go with what was req_d.
//
// c3_assert( god_u->dun_d == sav_u->req_d );
c3_assert( log_u->com_d >= god_u->dun_d );
{
u3_noun mat = u3ke_jam(u3nc(c3__save, u3i_chubs(1, &god_u->dun_d)));
// TODO: Verify change to req_d is correct during code review.
//
u3_noun mat = u3ke_jam(u3nc(c3__save, u3i_chubs(1, &sav_u->req_d)));
u3_newt_write(&god_u->inn_u, mat, 0);
// XX wait on some report of success before updating?
@ -918,7 +926,8 @@ _pier_work_release(u3_writ* wit_u)
if ( (0 == pir_u->ent_u) &&
(wit_u->evt_d < log_u->com_d) )
{
_pier_disk_load_commit(pir_u, (1ULL + god_u->dun_d), 1000ULL);
//_pier_disk_load_commit(pir_u, (1ULL + god_u->dun_d), 1000ULL);
_pier_db_load_commits(pir_u, (1ULL + god_u->dun_d), 1000ULL);
}
}
}