[HBASE-19350] TestMetaWithReplicas is flaky

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 1.4.0, 1.3.2, 2.0.0-beta-1, 2.0.0, 1.2.7
• Hadoop Flags: Reviewed
• Assignee: Chia-Ping Tsai
• Reporter: Chia-Ping Tsai
• Created: 27/Nov/17 08:42
• Updated: 01/Aug/18 06:21
• Resolved: 30/Nov/17 20:08

Description

If the size of RegionsInTransition is zero, the list passed to ClusterStatus will be null.

ClusterStatus.java

    Set<RegionState> rit = null;
    if (!proto.getRegionsInTransitionList().isEmpty()) {
      rit = new HashSet<RegionState>(proto.getRegionsInTransitionList().size());
      for (RegionInTransition region : proto.getRegionsInTransitionList()) {
        RegionState value = RegionState.convert(region.getRegionState());
        rit.add(value);
      }
    }

It causes NPE if someone try to do the for-each work. The HBaseFsckRepair is a real-life example.

HBaseFsckRepair.java

        for (RegionState rs: admin.getClusterStatus().getRegionsInTransition()) {
          if (rs.getRegion().equals(region)) {
            inTransition = true;
            break;
          }
        }

branch-2/master don't have this issue as the list of RegionsInTransition passed to ClusterStatus never be null.

ProtobufUtil.java

    List<RegionState> rit =
      new ArrayList<>(proto.getRegionsInTransitionList().size());
    for (RegionInTransition region : proto.getRegionsInTransitionList()) {
      RegionState value = RegionState.convert(region.getRegionState());
      rit.add(value);
}

Issue Links

• is related to: HBASE-19840 Flakey TestMetaWithReplicas