apple/foundationdb
1
0
Fork 0
mirror of https://github.com/apple/foundationdb.git synced 2025-06-01 18:56:00 +08:00
Code Issues Packages Projects Releases Wiki Activity

Fix timeout issue when using >3 coordinators

Browse Source 
The calculation to determine how many non-timeout replies had been
received was incorrect, causing rollback/rollforward requests to not be
sent, causing the dynamic knob subsystem to get stuck.
This commit is contained in:
Lukas Joswiak 2022-02-09 10:28:48 -08:00
parent 7fc4f0d649
commit e8354d82bd
1 changed files with 6 additions and 1 deletions

7
fdbserver/PaxosConfigConsumer.actor.cpp

@ -22,6 +22,7 @@
#include <algorithm>
#include <map>
#include <numeric>
#include "fdbserver/Knobs.h"
#include "flow/actorcompiler.h" // This must be the last #include.
@ -171,7 +172,11 @@ class GetCommittedVersionQuorum {
throw;
} else if (self->totalRepliesReceived == self->cfis.size() && self->quorumVersion.canBeSet() &&
!self->quorumVersion.isError()) {
if (self->replies.size() >= self->cfis.size() / 2 + 1) {
size_t nonTimeoutReplies =
std::accumulate(self->replies.begin(), self->replies.end(), 0, [](int value, auto const& p) {
return value + p.second.size();
});
if (nonTimeoutReplies >= self->cfis.size() / 2 + 1) {
// Make sure to trigger the quorumVersion if a timeout
// occurred, a quorum disagree on the committed version, and
// there are no more incoming responses. Note that this means