Apache
/
hbase
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

HBASE-18493 [AMv2] Skipped re-assignment of regions on crashed server through AssignmentManager.checkIfShouldMoveSystemRegionAsync() as those regions are handled by ServerCrashProcedure 

Signed-off-by: Michael Stack <stack@apache.org>
This commit is contained in:
Umesh Agashe 2017-08-15 14:00:04 -07:00 committed by Michael Stack
parent 4c74a73d57
commit acf9b87dca
1 changed files with 9 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
hbase-server/src/main/java/org/apache/hadoop/hbase/master/assignment/AssignmentManager.java
View File

@ -480,6 +480,15 @@ public class AssignmentManager implements ServerListener {
synchronized (checkIfShouldMoveSystemRegionLock) {
List<RegionPlan> plans = new ArrayList<>();
for (ServerName server : getExcludedServersForSystemTable()) {
if (master.getServerManager().isServerDead(server)) {
// TODO: See HBASE-18494 and HBASE-18495. Though getExcludedServersForSystemTable()
// considers only online servers, the server could be queued for dead server
// processing. As region assignments for crashed server is handled by
// ServerCrashProcedure, do NOT handle them here. The goal is to handle this through
// regular flow of LoadBalancer as a favored node and not to have this special
// handling.
continue;
}
List<HRegionInfo> regionsShouldMove = getCarryingSystemTables(server);
if (!regionsShouldMove.isEmpty()) {
for (HRegionInfo regionInfo : regionsShouldMove) {