Uploaded image for project: 'Core Server'
  1. Core Server
  2. SERVER-60751

move_chunk_critical_section_non_internal_client_abort.js does not consider config server stepdowns

    • Fully Compatible
    • ALL
    • v5.1, v5.0
    • Sharding EMEA 2021-11-15
    • 35

      After SERVER-47324 the collMod command requested to a shard directly will cause any ongoing migration to be aborted, in order to test this move_chunk_critical_section_non_internal_client_abort.js relies that a moveChunk command issued by a user will fail if executed with a concurrent collMod, however, in the sharding_csrs_continuous_config_stepdown suite we might find the following scenario:

      1. A router sends the _configsvrMoveChunk command to a config server
      2. The config server sends the moveChunk command to a shard
      3. The move chunk waits in the moveChunkHangAtStep5 failpoint (before committing the migration on the config server)
      4. We execute the collMod command directly on the shard
      5. The current migration is aborted
      6. The config server steps down before finishing the command
      7. The new config server primary retries the command

      Causing the test to fail, even though we're successfully aborting the migration on the donor shard at point 5 as expected. We should blacklist this test from the sharding_csrs_continuous_config_stepdown suite.

            marcos.grillo@mongodb.com Marcos José Grillo Ramirez
            marcos.grillo@mongodb.com Marcos José Grillo Ramirez
            0 Vote for this issue
            2 Start watching this issue