Details
-
Bug
-
Resolution: Fixed
-
Major
-
7.0.3, 7.1.0
-
1
-
Yes
-
KV Post-Neo (April)
Description
This is a production bug rather than test issue. We are not updating numOpsGet anymore for the gets made in this test (gets made via EventuallyPersistentEngine::getInner()). If the WorkLoadMonitor runs after the sets during the test and before all of the sets then it's possible to change the workload to write_heavy and not change it back. This is a regression introduced by this change https://review.couchbase.org/c/kv_engine/+/139024/7/engines/ep/src/ep_engine.cc#b2417.
Original manifestation:
http://cv.jenkins.couchbase.com/job/kv_engine.linux-CE/job/master/16363/
Failure of an engine_testapp test:
|
Running [0103/0159]: test multiple vb compactions with workload...Exceeded maximum wait time of 60000000us waiting for stat 'ep_workload_pattern' to be read_heavy (last value:write_heavy) - aborting.
|
Attachments
For Gerrit Dashboard: MB-50316 | ||||||
---|---|---|---|---|---|---|
# | Subject | Branch | Project | Status | CR | V |
173439,2 | MB-50316: Update numOpsGet on successful get via getInner() | master | kv_engine | Status: MERGED | +2 | +1 |