[SERVER-38142] Transaction coordinator should send prepare, commit, and abort directly to its local TransactionParticipant rather than over the network, take 2 Created: 14/Nov/18 Updated: 29/Oct/23 Resolved: 16/Jan/19 |
|
| Status: | Closed |
| Project: | Core Server |
| Component/s: | Sharding |
| Affects Version/s: | None |
| Fix Version/s: | 4.1.7 |
| Type: | Task | Priority: | Major - P3 |
| Reporter: | Esha Maharishi (Inactive) | Assignee: | Esha Maharishi (Inactive) |
| Resolution: | Fixed | Votes: | 0 |
| Labels: | ShardedTxn:DistributedCommit | ||
| Remaining Estimate: | Not Specified | ||
| Time Spent: | Not Specified | ||
| Original Estimate: | Not Specified | ||
| Attachments: |
|
| Backwards Compatibility: | Fully Compatible |
| Sprint: | Sharding 2018-12-31, Sharding 2019-01-14, Sharding 2019-01-28 |
| Participants: |
| Description |
|
Otherwise, in a split brained coordinator shard, the old coordinator primary might receive recoverCommit and find its local participant in prepare, because the new coordinator primary received coordinateCommit, had a stale ReplicaSetMonitor for itself, and sent prepareTransaction over the network to the old coordinator primary. |
| Comments |
| Comment by Esha Maharishi (Inactive) [ 16/Jan/19 ] |
|
Author: {'username': 'EshaMaharishi', 'email': 'esha.maharishi@mongodb.com', 'name': 'Esha Maharishi'}Message: |
| Comment by Esha Maharishi (Inactive) [ 14/Dec/18 ] |
|
Note that the participant on the old primary will be visible in the "prepared" state because prepareTransaction waits for writeConcern after checking the session back in. |