Details
-
Improvement
-
Resolution: Done
-
Major - P3
-
None
-
None
-
None
-
Storage Execution
Description
We currently use locks on resourceCappedInFlightForOtherDb and resourceCappedInFlightForLocalDb to ensure that snapshots see a consistent point in time. Now that we single-thread all writes to non-oplog capped collections, this is probably unnecessary overhead. It is likely better (and still correct) to just lock the oplog in MODE_S while taking the snapshot. Part of this work will be to verify that this is the case.