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

HBASE-26976 Update related comments after HMaster can load the live RS infos from local region (#4375) 

Signed-off-by: Duo Zhang <zhangduo@apache.org>
Signed-off-by: Yu Li <liyu@apache.org>
This commit is contained in:
LiangJun He 2022-04-30 10:53:13 +08:00 committed by GitHub
parent da37896588
commit b0c2832b6e
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
2 changed files with 5 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
hbase-server/src/main/java/org/apache/hadoop/hbase/master/HMaster.java
View File

@ -865,8 +865,8 @@ public class HMaster extends HBaseServerBase<MasterRpcServices> implements Maste
* from meta region</li>
* <li>Start region server tracker, construct the online servers set and find out dead servers and
* schedule SCP for them. The online servers will be constructed by scanning zk, and we will also
* scan the wal directory to find out possible live region servers, and the differences between
* these two sets are the dead servers</li>
* scan the wal directory and load from master local region to find out possible live region servers,
* and the differences between these two sets are the dead servers</li>
* </ol>
* </li>
* <li>If this is a new deploy, schedule a InitMetaProcedure to initialize meta</li>
@ -941,8 +941,6 @@ public class HMaster extends HBaseServerBase<MasterRpcServices> implements Maste
this.splitWALManager = new SplitWALManager(this);
}
tryMigrateMetaLocationsFromZooKeeper();
createProcedureExecutor();
@ -963,8 +961,8 @@ public class HMaster extends HBaseServerBase<MasterRpcServices> implements Maste
this.assignmentManager.setupRIT(ritList);
// Start RegionServerTracker with listing of servers found with exiting SCPs -- these should
// be registered in the deadServers set -- and with the list of servernames out on the
// filesystem that COULD BE 'alive' (we'll schedule SCPs for each and let SCP figure it out).
// be registered in the deadServers set -- and the servernames loaded from the WAL directory
// and master local region that COULD BE 'alive'(we'll schedule SCPs for each and let SCP figure it out).
// We also pass dirs that are already 'splitting'... so we can do some checks down in tracker.
// TODO: Generate the splitting and live Set in one pass instead of two as we currently do.
this.regionServerTracker.upgrade(

2
hbase-server/src/main/java/org/apache/hadoop/hbase/master/RegionServerTracker.java
View File

@ -129,7 +129,7 @@ public class RegionServerTracker extends ZKListener {
// '-SPLITTING'. Each splitting server should have a corresponding SCP. Log if not.
splittingServersFromWALDir.stream().filter(s -> !deadServersFromPE.contains(s)).
forEach(s -> LOG.error("{} has no matching ServerCrashProcedure", s));
// create ServerNode for all possible live servers from wal directory
// create ServerNode for all possible live servers from wal directory and master local region
liveServersBeforeRestart
.forEach(sn -> server.getAssignmentManager().getRegionStates().getOrCreateServer(sn));
ServerManager serverManager = server.getServerManager();