Ted Yu created HBASE-19813:
------------------------------
Summary: clone_snapshot fails with region failing to open when RS
group feature is enabled
Key: HBASE-19813
URL: https://issues.apache.org/jira/browse/HBASE-19813
Project: HBase
Issue Type: Bug
Environment: The following scenario came from support case.
In cluster 1, create RS group rsg. Move table to rsg group.
Take snapshot of the table and copy the snapshot to cluster 2 where there is no
group called rsg.
Cloning snapshot to table new_t4 on cluster 2 fails :
{code}
2018-01-09 11:45:30,468 INFO [RestoreSnapshot-pool68-t1] regionserver.HRegion:
Closed new_t4,,1514454789243.a6173d2955182ac5bde208301681c6af.
2018-01-09 11:45:30,468 INFO [MASTER_TABLE_OPERATIONS-shubh1-1:16000-0]
snapshot.CloneSnapshotHandler: Clone snapshot=snap_t3 on table=new_t4 completed!
2018-01-09 11:45:30,492 INFO [MASTER_TABLE_OPERATIONS-shubh1-1:16000-0]
hbase.MetaTableAccessor: Added 1
2018-01-09 11:45:30,492 WARN [MASTER_TABLE_OPERATIONS-shubh1-1:16000-0]
rsgroup.RSGroupBasedLoadBalancer: Group for table new_t4 is null
2018-01-09 11:45:30,492 DEBUG [MASTER_TABLE_OPERATIONS-shubh1-1:16000-0]
rsgroup.RSGroupBasedLoadBalancer: Group Information found to be null. Some
regions might be unassigned.
2018-01-09 11:45:30,492 WARN [MASTER_TABLE_OPERATIONS-shubh1-1:16000-0]
master.RegionStates: Failed to open/close a6173d2955182ac5bde208301681c6af on
null, set to FAILED_OPEN
{code}
Here is related code from RSGroupBasedLoadBalancer:
{code}
List<ServerName> candidateList = filterOfflineServers(info, servers);
for (RegionInfo region : regionList) {
currentAssignmentMap.put(region, regions.get(region));
}
if(candidateList.size() > 0) {
assignments.putAll(this.internalBalancer.retainAssignment(
currentAssignmentMap, candidateList));
{code}
candidateList is empty for table new_t4, leaving region for the table in
FAILED_OPEN state.
Reporter: Ted Yu
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
