Uploaded image for project: 'Core Server'
  1. Core Server
  2. SERVER-46516

Majority write concern is blocked by dropping snapshot on reconfig

XMLWordPrintableJSON

    • Replication
    • ALL
    • Hide

      Revert the change of write concern for setFCV and the check of config content change before dropping snapshot in SERVER-46288, then jstests/core/txns/await_prepared_transactions_on_FCV_downgrade.js will time out. max_time_ms_sharded_new_commands.js in sharding_csrs_continuous_config_stepdown will fail as well.

      Show
      Revert the change of write concern for setFCV and the check of config content change before dropping snapshot in  SERVER-46288 , then jstests/core/txns/await_prepared_transactions_on_FCV_downgrade.js  will time out. max_time_ms_sharded_new_commands.js  in sharding_csrs_continuous_config_stepdown will fail as well.
    • Repl 2020-03-23, Repl 2020-04-06
    • 60

      On reconfig, the primary drops all snapshots used to serve majority reads since the quorum may change by reconfig. The _currentCommittedSnapshot should be recalculated very soon, after a round of heartbeats in the worst case. However it's not the case. Majority writes are blocked until they are interrupted.

      We need to confirm that the _currentCommittedSnapshot is updated correctly after dropping snapshots and the waiters are signaled correctly.

        1. upgrade_downgrade_idempotency.js
          3 kB

            Assignee:
            backlog-server-repl [DO NOT USE] Backlog - Replication Team
            Reporter:
            siyuan.zhou@mongodb.com Siyuan Zhou
            Votes:
            0 Vote for this issue
            Watchers:
            11 Start watching this issue

              Created:
              Updated:
              Resolved: