[SERVER-4519] WRITETOJOURNAL and WRITETODATAFILES should not block readers Created: 18/Dec/11 Updated: 06/Dec/22 Resolved: 22/Mar/17 |
|
| Status: | Closed |
| Project: | Core Server |
| Component/s: | MMAPv1, Storage |
| Affects Version/s: | None |
| Fix Version/s: | None |
| Type: | Improvement | Priority: | Major - P3 |
| Reporter: | Dwight Merriman | Assignee: | Backlog - Storage Execution Team |
| Resolution: | Done | Votes: | 0 |
| Labels: | None | ||
| Remaining Estimate: | Not Specified | ||
| Time Spent: | Not Specified | ||
| Original Estimate: | Not Specified | ||
| Assigned Teams: |
Storage Execution
|
| Participants: |
| Description |
|
we should verify this, have a test. historically they have, i believe - when writes queue up that is. however they need not we could simply sleep() a long time in the writing to data fiels or journals while certain activity underway and verify we can still do queries. |
| Comments |
| Comment by Geert Bosch [ 22/Mar/17 ] |
|
This should be closed now: actual flushing and journal writing is now done in separate threads, so reads can proceed in parallel. There are still some critical sections of course, but we're not going to improve over what we have now. |
| Comment by Dwight Merriman [ 18/Dec/11 ] |
|
also, if these are bottlnecking let's be sure there are ways to tell what is up – which may already be the case in the serverStatus().dur stats – when tested let's look at that and see if there is sufficient telemetry to be clear on these things |