[SERVER-9471] hold write lock between finishing a batch and updating the oplog Created: 25/Apr/13  Updated: 19/Feb/16  Resolved: 23/Feb/15

Status: Closed
Project: Core Server
Component/s: Performance, Replication
Affects Version/s: 2.4.3
Fix Version/s: 2.7.8

Type: Improvement Priority: Major - P3
Reporter: Eric Milkie Assignee: Unassigned
Resolution: Done Votes: 0
Labels: None
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original Estimate: Not Specified

Backwards Compatibility: Fully Compatible
Participants:

 Description   

Right now we drop the write lock between finishing a batch and updating the oplog (and notifying upstream that we've applied ops).
This allows queued readers to run, and will delay our reporting of our progress upstream. This may in turn needlessly delay write concern getLastError calls on the primary.



 Comments   
Comment by Eric Milkie [ 23/Feb/15 ]

We now stay in ParallelBatchWriterMode while writing the ops to the oplog, which keeps readers queued.

Generated at Thu Feb 08 03:20:30 UTC 2024 using Jira 9.7.1#970001-sha1:2222b88b221c4928ef0de3161136cc90c8356a66.