initializeGlobalShardingStateForMongoD has to be run on shardsvrs during initialization. Currently, this is triggered from the following places:
1) In initAndListen
2) In ShardServerOpObserver::onInserts when an insert of the shard identity document is observed.
For FCBIS however, this does not work: On a new member that will undergo initial sync, point (1) will not find the shard identity document at all. Later, point (2) will not be triggered because FCBIS does not perform inserts like logical initial sync would.
As a result, after FCBIS has completed, the shard is left with uninitialized sharding state.
- depends on
-
SERVER-65251 FCBIS doesn't call to the onInitialSyncComplete ReplicaSetAwareService hook
- Closed
-
SERVER-64627 Need general method to handle in-memory state after initial sync
- Closed
- is depended on by
-
SERVER-66016 Create passthroughs for FCBIS on live config sets
- Closed
- is related to
-
SERVER-58935 Turn File Copy Based Initial Sync on by default
- Backlog
- related to
-
SERVER-66082 Ensure that cluster parameters are applied in-memory during file-copy based initial sync
- Closed
-
SERVER-64627 Need general method to handle in-memory state after initial sync
- Closed
-
SERVER-64628 More testing of adding nodes to sharded sets
- Closed