Ensure consistent failpoints even with the MultiShardedClusterFixture

XMLWordPrintableJSON

    • Type: Improvement
    • Resolution: Unresolved
    • Priority: Major - P3
    • None
    • Affects Version/s: None
    • Component/s: None
    • None
    • Catalog and Routing
    • CAR Team 2026-03-02, CAR Team 2026-03-16
    • None
    • None
    • None
    • None
    • None
    • None
    • None

      SERVER-117145 introduced multi-router support to sharding test suites.

      The failpoint util added with SERVER-119161 presents a convenient option to set a given failpoint across all mongos and non-config mongod nodes in the test cluster. Existing relevant commands calls to configure failpoints were also converted to use this new util.

      There are test suites which use MultiShardedClusterFixture to create multiple clusters, each of which has multiple mongos-es of their own. Here is one example. However, even with SERVER-119161, the runCommand override in query_shape_hash_stability.js (used in all test suites that use the MultiShardedClusterFixture) is not aware that failpoint configurations must be broadcasted.

      Ensure the failpoint util introduced in SERVER-119161 is MultiShardedClusterFixture-aware, similar to this helper in the hash stability overrides.

       

      As a side note, SERVER-120020 makes it is an error to try to set failpoints on a MultiRouterMongo connection. The MultiRouterMongo class is returned by the connect() method of the legacy shell for connection strings with multiple mongos-es, but the runCommand override instantiates the regular Mongo class explicitly, instead of obtaining a copy be calling connect(), so it is not affected by this change.

            Assignee:
            Enrico Golfieri
            Reporter:
            Kartal Kaan Bozdogan
            Votes:
            0 Vote for this issue
            Watchers:
            1 Start watching this issue

              Created:
              Updated: