Skip to content

test: stabilize rolling grain directory upgrade route publication #10903

Description

@ReubenBond

The Ubuntu net8.0 Functional job intermittently failed GrainDirectoryRollingUpgradeTests.RollingUpgrade_LocalToDistributed_NoErrors because the test captured an error log while gracefully removing old local-directory silos.

The logged operation was ClientDirectory publishing OnUpdateClientRoutes from one retiring silo to another retiring silo, which timed out after 30 seconds. The upgrade completed and all final directory invariants passed, but the unexpected error log caused Assert.Empty() to fail.

Observed in CI: https://github.com/dotnet/orleans/actions/runs/33132684117/job/98725565544

The exact test passed 5 consecutive local net8.0 runs after rebasing PR #10074 onto current main. Investigate shutdown ordering or expected exception classification for client-route publication during rolling removal so graceful topology changes do not intermittently emit an error.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions