[Colocate Table] Fix colocate table balance forever (#2744)
This commit is contained in:
@ -149,8 +149,8 @@ public class ColocateTableBalancer extends MasterDaemon {
|
||||
// But in previous version we had a bug that replicas of a tablet may be located on same host.
|
||||
// we have to check it.
|
||||
List<List<Long>> backendsPerBucketsSeq = colocateIndex.getBackendsPerBucketSeq(groupId);
|
||||
Set<String> hosts = Sets.newHashSet();
|
||||
OUT: for (List<Long> backendIds : backendsPerBucketsSeq) {
|
||||
Set<String> hosts = Sets.newHashSet();
|
||||
for (Long beId : backendIds) {
|
||||
Backend be = infoService.getBackend(beId);
|
||||
if (be == null) {
|
||||
@ -171,7 +171,6 @@ public class ColocateTableBalancer extends MasterDaemon {
|
||||
continue;
|
||||
}
|
||||
}
|
||||
Preconditions.checkState(unavailableBeId != -1);
|
||||
|
||||
// find the first bucket which contains the unavailable backend
|
||||
LOG.info("backend {} is unavailable in colocate group {}", unavailableBeId, groupId);
|
||||
|
||||
Reference in New Issue
Block a user