Quarantine RedisHubLifetimeManagerTests.ClientDisconnectsWithoutCompletingClientResultOnSecondServer failing intermittently

[github-actions]

Failing Test(s)

Microsoft.AspNetCore.SignalR.StackExchangeRedis.Tests.RedisHubLifetimeManagerTests.ClientDisconnectsWithoutCompletingClientResultOnSecondServer

Error Message

Assert.Throws() Failure: Exception type was not an exact match
Expected: typeof(Microsoft.AspNetCore.SignalR.HubException)
Actual:   typeof(System.IO.IOException)

Stacktrace

Details

Assert.Throws() Failure: Exception type was not an exact match
Expected: typeof(Microsoft.AspNetCore.SignalR.HubException)
Actual:   typeof(System.IO.IOException)
   at Xunit.Assert.Throws[T](Func`1 testCode)
   at Microsoft.AspNetCore.SignalR.StackExchangeRedis.Tests.RedisHubLifetimeManagerTests.ClientDisconnectsWithoutCompletingClientResultOnSecondServer()

Logs

Details

See build logs for detailed output. When a client disconnects while a client result is pending on the second server, the test expects a HubException but intermittently gets an IOException instead.

Build

https://dev.azure.com/dnceng-public/public/_build/results?buildId=1333947

Generated by Daily Test Quarantine Management · ◷

[github-actions]

Investigation

The test ClientDisconnectsWithoutCompletingClientResultOnSecondServer intermittently fails because the actual exception thrown is System.IO.IOException instead of the expected Microsoft.AspNetCore.SignalR.HubException.

Root cause: This is a race condition in the Redis backplane client result handling when a client disconnects unexpectedly on a second server. The test:

1. Invokes a client result from server2 to client1 (connected to server1)
2. Disconnects client1 before the result is completed
3. Expects a HubException indicating the client disconnected

When the Redis connection between the two servers drops or when the cleanup races with the client result wait, the SignalR infrastructure may throw an IOException (from the underlying Redis connection or stream) instead of wrapping it in a HubException.

The fix would be to ensure that when a client disconnects mid-client-result-invocation, all network-level exceptions from the Redis backplane are caught and wrapped in a HubException before propagating to the caller.

Quarantine PR: The test is being quarantined to unblock CI while the exception wrapping issue is investigated.

Generated by Daily Test Quarantine Management · ◷

[github-actions]

Investigation

The test ClientDisconnectsWithoutCompletingClientResultOnSecondServer intermittently fails with:

Assert.Throws() Failure: Exception type was not an exact match
Expected: typeof(Microsoft.AspNetCore.SignalR.HubException)
Actual:   typeof(System.IO.IOException)
---- System.IO.IOException : Connection '0e9a7dbd-bed3-4098-ad6f-cc14559a158e' does not exist.

Root cause: There is a race condition in RedisHubLifetimeManager.InvokeConnectionAsync (RedisHubLifetimeManager.cs:403). The test:

1. Has client1 connected to server1/manager2
2. Invokes a client result from manager1 to connection1 (cross-server via Redis)
3. Disconnects client1 before the result is received

When the timing is tight between the disconnect propagation and the Redis invocation, InvokeConnectionAsync throws a System.IO.IOException ("Connection does not exist") directly from the Redis layer rather than the expected HubException. The fix would be to ensure that InvokeConnectionAsync wraps or rethrows IOException as HubException consistently when the connection no longer exists.

This test is being quarantined while the fix is investigated.

Note

🔒 Integrity filter blocked 9 items

The following items were blocked because they don't meet the GitHub integrity level.

• #65928 pull_request_read: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• #65869 pull_request_read: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• #65893 pull_request_read: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• #65815 pull_request_read: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• EditForm: oninput loses typed characters during rapid typing / long key press on slow network #65994 list_issues: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• InteractiveAuto WebAssembly fails under FallbackPolicy RequireAuthenticatedUser #65993 list_issues: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• The "pattern" paramter of Microsoft.AspNetCore.Builder.EndpointRouteBuilderExtensions.MapFallback is ignored when EnableRequestDelegateGenerator is true #65992 list_issues: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• Perf improvement: httpsys-encoded-url #65988 list_issues: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• Publish Microsoft.AspNetCore.JsonPatch.SystemTextJson without ASP.NET Core Runtime dependency #65986 list_issues: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".

To allow these resources, lower min-integrity in your GitHub frontmatter:

tools:
  github:
    min-integrity: approved  # merged | approved | unapproved | none

Generated by Daily Test Quarantine Management · ◷