[SERVER-74113] Investigate Long/Medium snapshot scan regressions Created: 16/Feb/23 Updated: 29/Mar/23 |
|
| Status: | Open |
| Project: | Core Server |
| Component/s: | None |
| Affects Version/s: | None |
| Fix Version/s: | None |
| Type: | Task | Priority: | Major - P3 |
| Reporter: | Mihai Andrei | Assignee: | Backlog - Query Execution |
| Resolution: | Unresolved | Votes: | 0 |
| Labels: | None | ||
| Remaining Estimate: | Not Specified | ||
| Time Spent: | Not Specified | ||
| Original Estimate: | Not Specified | ||
| Assigned Teams: |
Query Execution
|
| Sprint: | QE 2023-03-06, QE 2023-03-20 |
| Participants: | |
| Story Points: | 3 |
| Description |
|
| Comments |
| Comment by Ivan Fefer [ 29/Mar/23 ] |
|
So this scans are from LLTMixed workload:
In perf reports most time is spent in WiredTired cursors. I looked for a difference in how classic and sbe handle cursors, but there is no significant changes there. The only thing was that classic creates cursors during first call of doWork and SBE creates cursors during open(). However in On that I am out of ideas and didn't have a change to investigate more. |
| Comment by David Storch [ 28/Mar/23 ] |
|
ivan.fefer@mongodb.com as we discussed in standup today, could you provide a status update on this ticket? What do you currently know about the root cause of these regressions? Do you have any potential fixes or experiments planned? |
| Comment by Ivan Fefer [ 03/Mar/23 ] |
|
perf reports show that the most time is spent in WiredTiger for this workload. Because there are a lot of very short queries and writes in this workload, I tried to create cursors as late as possible and close them as soon as possible, but it didn't show any statistically significant change. Looks like |