alexander.gorrod, based on SERVER-34082, I think cursor restart can happen fast enough that high level cursor operations like search can spin in a core: https://jira.mongodb.org/browse/SERVER-34082?focusedCommentId=1858240&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-1858240
It would be a relatively simple change to back off and yield the processor in the case of repeated restarts in high-level cursor operations.
- causes
-
WT-4049 Performance drop in long tests
- Closed
- related to
-
SERVER-34082 MongoDB stalls randomly
- Closed
-
WT-4139 rename the cursor restart statistic to match implementation
- Closed
- links to