Details
-
Bug
-
Status: Resolved
-
Major
-
Resolution: Fixed
-
None
-
None
-
None
Description
I see this in recent nightly fails on branch-2, my experimental branch. It happened in the two builds last night. It was only test to fail. Both failed with:
2020-04-10 07:42:46,700 WARN [Time-limited test] regionserver.MultiVersionConcurrencyControl(228): STUCK: MultiVersionConcurrencyControl
{readPoint=8684, writePoint=8686}I ran the test in my harness and it pushes up against heap limits for sure, especially after recent cut down in forked jvm size.
For now, see if alleviating scope of testWritesWhileScanning just when called from this test helps w/ flakeyness. Probably worth a deep dive why we get into this state w/ in-memory compaction.