Pt 2531 pt online schema change creates sleeping threads in replicas

[svetasmirnova]

• The contributed code is licensed under GPL v2.0
• Contributor Licence Agreement (CLA) is signed
• util/update-modules has been ran
• Documentation updated
• Test suite update

[Copilot]

🟡 Changes recommended

The new replica-refresh callback invocation and stale-replica disconnect logic introduce confirmed runtime/behavioral bugs (callback argument mismatch, undef deref, and incorrect disconnect behavior).

Once you've addressed the issues Copilot identified, you can request another Copilot review.

Pull request overview

This PR aims to prevent pt-online-schema-change from leaving sleeping (stale) connections on replicas by reusing existing replica connections across refreshes and disconnecting replicas that are no longer part of the refreshed replica set.

Changes:

• Pass the current replica list into replica refresh paths to enable connection reuse.
• Reuse known replica connections when reading replicas from a DSN table and disconnect stale replicas.
• Adjust the pt-2168 test timing and fix minor typos/formatting.

File summaries

File	Description
t/pt-online-schema-change/pt-2168.t	Tweaks replica-delay timing used by the test harness.
lib/ReplicaLagWaiter.pm	Changes how replica refresh callbacks are invoked during lag waiting.
lib/MasterSlave.pm	Reuses known replica connections when reading from DSN tables and attempts to disconnect stale replicas.
bin/pt-online-schema-change	Mirrors the library changes in the embedded code paths (replica refresh + DSN-table reuse + disconnect stale replicas) and adjusts stopped-replication behavior.

Review details

Suppressed comments (1)

lib/MasterSlave.pm:1142

• The stale-replica disconnect loop disconnects known_replica as soon as it finds any non-matching new_replica (i.e., it can disconnect a replica even if it exists later in @cxn). Also, $args{replicas} may be undef. Only disconnect after verifying no match against any new replica, and guard the dereference.

   # Close database connections in stale replicas
   REPLICAS:
   foreach my $known_replica ( @{$args{replicas}} ) {
      foreach my $new_replica ( @cxn ) {
         if ( $known_replica->{dsn}->{h} eq $new_replica->{dsn}->{h} and
            $known_replica->{dsn}->{P} eq $new_replica->{dsn}->{P} ) {
            next REPLICAS;
         }
         $known_replica->{dbh}->disconnect;
      }
   }

• Files reviewed: 3/4 changed files
• Comments generated: 2
• Review effort level: Lite

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[it-percona-cla]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you all sign our Contributor License Agreement before we can accept your contribution.
1 out of 2 committers have signed the CLA.

✅ svetasmirnova
❌ Copilot
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[matias-sanchez]

adding enhancements based on testing with 1 source and 4 replicas in the dsns table. With both applied i saw no reconnections over many chunks.