vitalif
/
vitastor
5
4
Fork 9
Code Issues 1 Pull Requests 2 Actions Releases 69 Wiki Activity

Only die when detecting a real race condition, not just a CAS failure

non-odp-rdma
Vitaliy Filippov 2022-01-05 17:05:25 +03:00
parent 68b6763ebe
commit 515a2e6e33
1 changed files with 6 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
src/osd_cluster.cpp
View File

@ -858,10 +858,13 @@ void osd_t::report_pg_states()
if (null_byte == 0) if (null_byte == 0)
{ {
auto pg_it = pgs.find({ .pool_id = pool_id, .pg_num = pg_num }); auto pg_it = pgs.find({ .pool_id = pool_id, .pg_num = pg_num });
if (pg_it != pgs.end() && pg_it->second.state != PG_OFFLINE && pg_it->second.state != PG_STARTING) if (pg_it != pgs.end() && pg_it->second.state != PG_OFFLINE && pg_it->second.state != PG_STARTING &&
kv.value["primary"].uint64_value() != 0 &&
kv.value["primary"].uint64_value() != this->osd_num)
{ {
// Live PG state update failed // PG is somehow captured by another OSD
printf("Failed to report state of pool %u PG %u which is live. Race condition detected, exiting\n", pool_id, pg_num); printf("BUG: OSD %lu captured our PG %u/%u. Race condition detected, exiting\n",
kv.value["primary"].uint64_value(), pool_id, pg_num);
force_stop(1); force_stop(1);
return; return;
} }