Apache9 commented on code in PR #5442:
URL: https://github.com/apache/hbase/pull/5442#discussion_r1349474487
##########
hbase-server/src/main/java/org/apache/hadoop/hbase/master/assignment/AssignmentManager.java:
##########
@@ -1398,27 +1398,37 @@ private void checkOnlineRegionsReport(ServerStateNode
serverNode, Set<byte[]> re
continue;
}
final long lag = 1000;
- regionNode.lock();
try {
- long diff = EnvironmentEdgeManager.currentTime() -
regionNode.getLastUpdate();
- if (regionNode.isInState(State.OPENING, State.OPEN)) {
- // This is possible as a region server has just closed a region but
the region server
- // report is generated before the closing, but arrive after the
closing. Make sure there
- // is some elapsed time so less false alarms.
- if (!regionNode.getRegionLocation().equals(serverName) && diff >
lag) {
- LOG.warn("Reporting {} server does not match {} (time since last "
- + "update={}ms); closing...", serverName, regionNode, diff);
- closeRegionSilently(serverNode.getServerName(), regionName);
- }
- } else if (!regionNode.isInState(State.CLOSING, State.SPLITTING)) {
- // So, we can get report that a region is CLOSED or SPLIT because a
heartbeat
- // came in at about same time as a region transition. Make sure
there is some
- // elapsed time so less false alarms.
- if (diff > lag) {
- LOG.warn("Reporting {} state does not match {} (time since last
update={}ms)",
- serverName, regionNode, diff);
+ // It is likely that another thread is currently holding the lock. To
avoid deadlock, use
+ // tryLock waiting for a specified period of time
+ if (regionNode.tryLock(10, TimeUnit.SECONDS)) {
Review Comment:
tryLock should be placed outside the try block, as we do not need to unlock
if we get a false here.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
