[SERVER-45642] A syncdelay value of 0 causes the WiredTigerCheckpointThread to run as often as possible Created: 17/Jan/20 Updated: 29/Oct/23 Resolved: 02/Jun/20 |
|
| Status: | Closed |
| Project: | Core Server |
| Component/s: | Storage |
| Affects Version/s: | None |
| Fix Version/s: | 4.4.0-rc9, 4.7.0 |
| Type: | Bug | Priority: | Major - P3 |
| Reporter: | Justin Seyster | Assignee: | Dianna Hohensee (Inactive) |
| Resolution: | Fixed | Votes: | 0 |
| Labels: | None | ||
| Remaining Estimate: | Not Specified | ||
| Time Spent: | Not Specified | ||
| Original Estimate: | Not Specified | ||
| Issue Links: |
|
||||||||||||
| Backwards Compatibility: | Fully Compatible | ||||||||||||
| Operating System: | ALL | ||||||||||||
| Backport Requested: |
v4.4
|
||||||||||||
| Sprint: | Execution Team 2020-06-01, Execution Team 2020-06-15 | ||||||||||||
| Participants: | |||||||||||||
| Description |
|
The help text for the --syncdelay command line argument (which populates wiredTigerGlobalOptions.checkpointDelaySecs) indicates that a value of 0 will stop the checkpoint thread from running at all. In the actual thread, though, it gets interpreted as a delay of 0, causing it to loop as fast as it can: |
| Comments |
| Comment by Githook User [ 05/Jun/20 ] |
|
Author: {'name': 'Dianna Hohensee', 'email': 'dianna.hohensee@mongodb.com', 'username': 'DiannaHohensee'}Message: Also move JS tests using high syncdelay values to using the now working 0 value to disable checkpointing. (cherry picked from commit 8ee572abe5a47d3d49e7436c1a376ac395517add) |
| Comment by Githook User [ 02/Jun/20 ] |
|
Author: {'name': 'Dianna Hohensee', 'email': 'dianna.hohensee@mongodb.com', 'username': 'DiannaHohensee'}Message: Also move JS tests using high syncdelay values to using the now working 0 value to disable checkpointing. |
| Comment by Connie Chen [ 23/Jan/20 ] |
|
We should make 0 disable checkpointing work for testing, but take out the documentation that mentions 0 sync delay . Also make sure the max value is reasonable for sync-delay |
| Comment by Justin Seyster [ 21/Jan/20 ] |
|
One option for testing would be to set the syncdelay to a very high value (higher than the timeout for any of the tests). |
| Comment by David Daly [ 21/Jan/20 ] |
|
I'm fine with changing documentation to reflect reality. If we do that can we add a knob that does what we originally expected for internal testing purposes? |
| Comment by Alexander Gorrod [ 19/Jan/20 ] |
|
Given that the behavior of disabling checkpoint has been broken since the first 3.6 release, I think we should consider changing the documentation to match current behavior. Disabling checkpoints has fairly far-reaching and sometimes surprising consequences, like:
|