Phoenix 7920 Add replay-side handling for direct DEGRADED_STANDBY → STANDBY_TO_ACTIVE transition (ANISTS → AISTS)#2573
Open
Himanshu-g81 wants to merge 13 commits into
Conversation
added 13 commits
July 11, 2026 16:27
… on direct failover
… (avoid epoch rewind)
Capture current time at the start of initializeLastRoundProcessed() (before the HA group state read and file scans) and thread it through all three init branches - DEGRADED_STANDBY, STANDBY_TO_ACTIVE, and the SYNC path - via a new fallback-time overload on the parent. This anchors the starting round to when init began, so a SYNC -> DEGRADED_STANDBY transition during init cannot push the round (and the SYNC-path recovery floor) past the degradation point. Also documents the zero lastSyncStateTimeInMs fallback on initLastRoundsFromLastSyncPoint.
…on test The zero-lastSyncStateTimeInMs guard in initLastRoundsFromLastSyncPoint is shared by the DEGRADED_STANDBY and STANDBY_TO_ACTIVE init paths, but only the STANDBY_TO_ACTIVE path had a zero-sync regression test. The guard also corrected the DEGRADED path (which previously computed round(min(0, frontier)) = round(0) = ReplicationRound(0,0), the epoch-rewind bug). This adds the symmetric DEGRADED_STANDBY test so a future refactor that diverges the two paths cannot silently reintroduce the epoch rewind. Test count 44 -> 45 (failsafe XML verified GREEN).
…ed cross-cluster helpers
…ly-driven replay, SYNC init
…advances sync point
…-aware race Cluster2 is peer-aware and auto-reacts to Cluster1's ACTIVE_NOT_IN_SYNC by driving its own LOCAL record to DEGRADED_STANDBY during store-and-forward entry. That reaction can still be in flight when transitionCluster2(DEGRADED_STANDBY) runs on a slow box, so setHAGroupStatusIfNeeded could throw InvalidClusterRoleTransitionException on a redundant same-state transition. Poll a bounded 10s window for the record to settle at the target before the no-op check so the transition is robust whether or not the reaction has landed.
…to fix flaky failover The promotion poll (driveReplayUntil ACTIVE_IN_SYNC) was flaky on slow/loaded boxes: it timed out at 120s on a 176.8s run but passed at ~57s runs. Root cause is not the guarded Fix apache#1 code but the test harness. Cluster1's writer was left running through Stage 3; its rotation thread emits an empty round .plog every round-duration into Cluster2's 'in' dir. That file perpetually occupies the [nextRoundToProcess, currentTimestampRound] window that shouldTriggerFailover() condition 4 (getNewFiles(...).isEmpty()) scans, so promotion could only land in a race window between the replay catching up and the next file arriving. The failed run shows 239/239 checks blocked with 'New files exist...'. Fix: close logGroup before the promotion drive, modeling the real failover precondition (the old ACTIVE is gone, producing no more logs). File production stops, the replay drains, condition 4 clears, and promotion is deterministic at any round duration. Batch B is already forwarded (Stage 2), so the SYNCED_RECOVERY rewind still re-replays it and zero RPO holds; close() is idempotent so afterTest's close is a no-op. Also drop the local design-spec path from the class javadoc (spec is not being pushed). Verified: Tests run: 1, Failures: 0, Time elapsed 57.61s; log shows promotion firing ~2s after writer close via 'Failover can be triggered'.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.