Mirroristas/etcd
2
0
Fork 0
mirror of https://github.com/etcd-io/etcd.git synced 2024-09-27 06:25:44 +00:00
Code Issues Packages Projects Releases Wiki Activity

etcdserver: added a summary for the CompactHashCheck method

Browse Source 
Signed-off-by: Benjamin Wang <wachao@vmware.com>
This commit is contained in:
Benjamin Wang 2022-11-26 05:52:45 +08:00
parent 85fc09d09b
commit e95e82f0b9
1 changed files with 36 additions and 21 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

57
server/etcdserver/corrupt.go
View File

@ -252,6 +252,17 @@ func (cm *corruptionChecker) PeriodicCheck() error {
return nil
}
// CompactHashCheck is based on the fact that 'compactions' are coordinated
// between raft members and performed at the same revision. For each compacted
// revision there is KV store hash computed and saved for some time.
//
// This method communicates with peers to find a recent common revision across
// members, and raises alarm if 2 or more members at the same compact revision
// have different hashes.
//
// We might miss opportunity to perform the check if the compaction is still
// ongoing on one of the members or it was unresponsive. In such situation the
// method still passes without raising alarm.
func (cm *corruptionChecker) CompactHashCheck() {
cm.lg.Info("starting compact hash check",
zap.String("local-member-id", cm.hasher.MemberId().String()),
@ -314,27 +325,7 @@ func (cm *corruptionChecker) checkPeerHashes(leaderHash mvcc.KeyValueHash, peers
// All members have the same CompactRevision and Hash.
if len(hash2members) == 1 {
if peersChecked == len(peers) {
cm.lg.Info("successfully checked hash on whole cluster",
zap.Int("number-of-peers-checked", peersChecked),
zap.Int64("revision", leaderHash.Revision),
zap.Int64("compactRevision", leaderHash.CompactRevision),
)
cm.mux.Lock()
if leaderHash.Revision > cm.latestRevisionChecked {
cm.latestRevisionChecked = leaderHash.Revision
}
cm.mux.Unlock()
return true
}
cm.lg.Warn("skipped revision in compaction hash check; was not able to check all peers",
zap.Int("number-of-peers-checked", peersChecked),
zap.Int("number-of-peers", len(peers)),
zap.Int64("revision", leaderHash.Revision),
zap.Int64("compactRevision", leaderHash.CompactRevision),
)
// The only case which needs to check next hash
return false
return cm.handleConsistentHash(leaderHash, peersChecked, len(peers))
}
// Detected hashes mismatch
@ -407,6 +398,30 @@ func (cm *corruptionChecker) checkPeerHashes(leaderHash mvcc.KeyValueHash, peers
return true
}
func (cm *corruptionChecker) handleConsistentHash(hash mvcc.KeyValueHash, peersChecked, peerCnt int) bool {
if peersChecked == peerCnt {
cm.lg.Info("successfully checked hash on whole cluster",
zap.Int("number-of-peers-checked", peersChecked),
zap.Int64("revision", hash.Revision),
zap.Int64("compactRevision", hash.CompactRevision),
)
cm.mux.Lock()
if hash.Revision > cm.latestRevisionChecked {
cm.latestRevisionChecked = hash.Revision
}
cm.mux.Unlock()
return true
}
cm.lg.Warn("skipped revision in compaction hash check; was not able to check all peers",
zap.Int("number-of-peers-checked", peersChecked),
zap.Int("number-of-peers", peerCnt),
zap.Int64("revision", hash.Revision),
zap.Int64("compactRevision", hash.CompactRevision),
)
// The only case which needs to check next hash
return false
}
func (cm *corruptionChecker) uncheckedRevisions() []mvcc.KeyValueHash {
cm.mux.RLock()
lastRevisionChecked := cm.latestRevisionChecked