[imbajin]

Before → After

Before: stale persisted SHARDGROUP metadata could omit a migrated Store and exclude its partition before JRaft restored the durable configuration.

After: validate PD membership and persisted graph partitions, then refresh derived metadata once per group only when its engine is absent. Preserve running engines and durable Raft configuration; missing, invalid, or unavailable PD metadata blocks recovery and retains local data.

Validation boundary

The source PR reports 19 tests with no failures or skips on Java 17, including a real JRaft node reopening a persisted one-voter configuration without replacing it with a different PD bootstrap. This does not claim independent multi-node migration/restart acceptance.

Source PR: 🔗 hugegraph/hugegraph #344

Visual overview

Direct PD partition NOT_FOUND skips deleted-graph recovery while retaining local data; surviving graphs still recover. Other metadata failures remain fail-closed. Durable Raft configuration stays authoritative, and empty-log bootstrap preserves learner roles. No independent multi-node migration or RC acceptance is claimed.

[codecov]

Codecov Report

❌ Patch coverage is 90.54054% with 7 lines in your changes missing coverage. Please review.
✅ Project coverage is 57.57%. Comparing base (608cb1d) to head (5711436).
⚠️ Report is 5 commits behind head on master.

Files with missing lines	Patch %	Lines
.../apache/hugegraph/store/meta/PartitionManager.java	93.84%	1 Missing and 3 partials ⚠️
.../java/org/apache/hugegraph/pd/client/PDClient.java	83.33%	1 Missing ⚠️
...g/apache/hugegraph/store/pd/DefaultPdProvider.java	50.00%	0 Missing and 1 partial ⚠️
...java/org/apache/hugegraph/store/pd/PdProvider.java	0.00%	1 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff              @@
##             master    #3417      +/-   ##
============================================
+ Coverage     57.27%   57.57%   +0.30%     
+ Complexity     6577     2479    -4098     
============================================
  Files           776      776              
  Lines         68881    68962      +81     
  Branches       9431     9453      +22     
============================================
+ Hits          39449    39708     +259     
+ Misses        25804    25613     -191     
- Partials       3628     3641      +13     

Flag	Coverage Δ		*Carryforward flag
commons	96.37% <ø> (ø)		Carriedforward from f005ef6
hstore	36.11% <ø> (+0.02%)	⬆️
pd	52.34% <50.00%> (+0.18%)	⬆️
server_hbase	42.18% <ø> (ø)		Carriedforward from f005ef6
server_memory	55.98% <ø> (ø)		Carriedforward from f005ef6
server_rocksdb	42.65% <ø> (ø)		Carriedforward from f005ef6
store	35.00% <90.54%> (+1.85%)	⬆️

*This pull request uses carry forward flags. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[bitflicker64]

Blocking: yes. Summary: The direct PD shard group refresh, the running-engine guard and the role-aware /learner peers match the PR goal, and the earlier learner finding is fixed at this head. Startup now exits the Store when PD no longer has a persisted graph partition, where the base skipped that partition.

[bitflicker64]

Blocking: no. Score: 9/10. Summary: Recovery validates PD membership and persisted partitions before refreshing derived SHARDGROUP metadata, keeps durable Raft configuration and running engines authoritative, and both earlier findings (learner bootstrap, dropped-graph NOT_FOUND) are fixed with tests that run in CI.