Rework SimpleSetupCluster in cluster tests, and longer timeouts for select tests

[kevin-montrose]

Reworks ClusterTestUtils.SimpleSetupCluster to be async and to pause and very state between each step - this increases reliability and help with diagnosing failures as we know the cluster is in a stable state before beginning the "real" test.

Also increases timeouts (to 2 minutes) for RepeatedCreateDeleteAsync, MigrateVectorSetWhileModifyingAsync, and MigrateVectorStressAsync - these tests occasionally take nearly 1 minute in Debug builds on CI machines, presumably some failures just legitimately ran out of time.

As part of increasing timeouts, makes ClusterTestContext aware of [CancelAfter] and uses its value during setup if available.

TODO:

• dev version ([DEV] Rework SimpleSetupCluster in cluster tests, and longer timeouts for select tests #1763)

[Copilot]

Pull request overview

This PR improves cluster test reliability by making cluster setup more deterministic/async-aware and by extending timeouts for a few long-running cluster/vector migration tests. It also adds additional “wait until stable” checks during SimpleSetupCluster and updates several test helpers to use async StackExchange.Redis APIs.

Changes:

• Reworked ClusterTestUtils.SimpleSetupCluster into an async implementation with additional stabilization checks between meet/slot/replication steps.
• Updated multiple cluster tests to use async setup and increased timeouts (via [CancelAfter(120_000)]) for select slow/flaky tests.
• Converted several ClusterTestUtils helpers (connect/reconnect, nodes parsing, replication polling) to async variants and updated callsites.

Reviewed changes

Copilot reviewed 8 out of 8 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
test/Garnet.test.cluster/VectorSets/ClusterVectorSetTests.cs	Converts several tests and cluster setup wrapper to async; adds longer [CancelAfter] timeouts for select tests.
test/Garnet.test.cluster/ReplicationTests/ClusterReplicationBaseTests.cs	Converts select tests/helpers to async waiting APIs (e.g., connected replica polling).
test/Garnet.test.cluster/ClusterTestUtils.cs	Core change: async SimpleSetupClusterAsync, new async helpers, and additional stabilization checks during cluster formation.
test/Garnet.test.cluster/ClusterTestContext.cs	Uses NUnit timeout metadata to size the context CTS; makes replica attach/sync helper async-aware.
test/Garnet.test.cluster/ClusterNegativeTests.cs	Updates tests to await async connected-replica polling.
test/Garnet.test.cluster/ClusterMigrateTests.cs	Updates one test to await async “pick any other node” helper.
libs/cluster/Session/RespClusterBasicCommands.cs	Wraps CLUSTER GOSSIP handling in try/catch with logging; retains epoch-release pattern around merge.
libs/cluster/Server/Replication/ReplicationManager.cs	Makes logging null-safe (logger?.) in resync path.

Comments suppressed due to low confidence (2)

test/Garnet.test.cluster/ClusterTestUtils.cs:2348

• ClusterNodesAsync(IPEndPoint ...) is marked async but doesn't await anything and just calls the synchronous server.ClusterNodes(). With TreatWarningsAsErrors=true, this will produce CS1998 and fail the build. Either call/await the real async API (e.g., await server.ClusterNodesAsync() if available) or remove async and return a completed Task from the synchronous result.

        public async Task<ClusterConfiguration> ClusterNodesAsync(IPEndPoint endPoint, ILogger logger = null)
        {
            try
            {
                var server = redis.GetServer(endPoint);
                return server.ClusterNodes();
            }

test/Garnet.test.cluster/ClusterTestContext.cs:813

• The loops that wait for replica recovery/AOF sync and validate replica data use for (var i = primary_count; i < replica_count; i++), but replica_count is a count, not the upper node index. For primary_count=1, replica_count=1 this loop never runs, so the method can return without waiting/validating any replicas. Use i < primary_count + replica_count (matching the earlier loops in this method) for the recovery/AOF sync and validation loops as well.

            // Wait for recovery and AofSync
            for (var i = primary_count; i < replica_count; i++)
            {
                clusterTestUtils.WaitForReplicaRecovery(i, logger);
                clusterTestUtils.WaitForReplicaAofSync(0, i, logger);
            }

            await clusterTestUtils.WaitForConnectedReplicaCountAsync(0, replica_count, logger: logger).ConfigureAwait(false);

            // Validate data on replicas
            for (var i = primary_count; i < replica_count; i++)
            {

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.