[SERVER-35657] Do not delay journal flushes when operations are waiting for oplog visibility Created: 18/Jun/18 Updated: 29/Oct/23 Resolved: 24/Aug/18 |
|
| Status: | Closed |
| Project: | Core Server |
| Component/s: | Storage |
| Affects Version/s: | 3.6.6, 4.0.0-rc5 |
| Fix Version/s: | 3.6.9, 4.0.3, 4.1.3 |
| Type: | Bug | Priority: | Major - P3 |
| Reporter: | Cristopher Stauffer | Assignee: | Louis Williams |
| Resolution: | Fixed | Votes: | 0 |
| Labels: | bkp | ||
| Remaining Estimate: | Not Specified | ||
| Time Spent: | Not Specified | ||
| Original Estimate: | Not Specified | ||
| Issue Links: |
|
||||||||||||
| Backwards Compatibility: | Fully Compatible | ||||||||||||
| Operating System: | ALL | ||||||||||||
| Backport Requested: |
v4.0, v3.6
|
||||||||||||
| Steps To Reproduce: | 1. Configure 1-node replica set |
||||||||||||
| Sprint: | Storage NYC 2018-07-02, Storage NYC 2018-07-16, Storage NYC 2018-07-30, Storage NYC 2018-08-13, Storage NYC 2018-08-27 | ||||||||||||
| Participants: | |||||||||||||
| Linked BF Score: | 0 | ||||||||||||
| Description |
|
In a recent examination of transactions performance it was observed that specific combinations of read concern, topology, and causal settings caused significant performance slowdowns for read operations. Most notably the single-node replica sets using local read concern and causal consistency. In a build failure ticket we tracked this to the following work: |
| Comments |
| Comment by Githook User [ 18/Sep/18 ] |
|
Author: {'name': 'Louis Williams', 'email': 'louis.williams@mongodb.com', 'username': 'louiswilliams'}Message: (cherry picked from commit fa0ed40e73e65a1d8715a2dcd0be22f254787bb0) Conflicts: |
| Comment by Githook User [ 17/Sep/18 ] |
|
Author: {'name': 'Louis Williams', 'email': 'louis.williams@mongodb.com', 'username': 'louiswilliams'}Message: (cherry picked from commit fa0ed40e73e65a1d8715a2dcd0be22f254787bb0) |
| Comment by Githook User [ 24/Aug/18 ] |
|
Author: {'name': 'Louis Williams', 'email': 'louis.williams@mongodb.com', 'username': 'louiswilliams'}Message: |
| Comment by Louis Williams [ 22/Aug/18 ] |
|
A fix for this regression should be to include waiters of _opsBecameVisibleCV when deciding whether or not delay journal flushes. |