[HBASE-21562] TestRestoreSnapshotFromClientAfterSplittingRegions and related tests are flakey

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Duplicate
• Fix Version/s: 2.1.2
• Component/s: test
• Assignee: Unassigned
• Reporter: Michael Stack
• Created: 06/Dec/18 19:02
• Updated: 07/Dec/18 04:48
• Resolved: 07/Dec/18 04:48

Description

Fails 60% of the time on GCE runs. Messes up our nightlies for branch-2.1 and branch-2.0 at least.

Looking its a bit tough figuring what is going on. Test asks us split regions. The split starts then hangs. Last thing reported is:

2018-12-06 10:20:30,823 INFO [PEWorker-16] procedure.MasterProcedureScheduler(741): Took xlock for pid=174, state=RUNNABLE:SPLIT_TABLE_REGION_PREPARE; SplitTableRegionProcedure table=testRestoreSnapshotAfterSplittingRegions_1_regionReplication_3-1544120421990, parent=034bb3ebb3f9a7442f927caacdda5354, daughterA=fbe392ca659b3913181d05ac4fb19b4c, daughterB=3646ac333722af33c32e6f3428d23f95

... then all we get is that the worker is stuck.