> On Feb. 28, 2017, 6:32 p.m., Alejandro Fernandez wrote: > >
Thanks Alejanro, i wasn't ware of that; I'll do the modification to the other approach then. > On Feb. 28, 2017, 6:32 p.m., Alejandro Fernandez wrote: > > ambari-server/src/main/resources/common-services/HDFS/2.1.0.2.0/metainfo.xml > > Lines 64 (patched) > > <https://reviews.apache.org/r/57147/diff/1/?file=1651734#file1651734line64> > > > > What happens if a rebalance command is issued and one is already > > running? > > > > I don't want a process running on Ambari for up to 2 hours. And for > > very large clusters (1000+ nodes), it may still timeout. > > Alejandro Fernandez wrote: > HDFS stores a file in /system/balancer.id while the balancer is on-going. > As soon as it finishes, it is deleted. > You can use that to make the call return right away and run in the > background, and also prevent subsequent calls from launching another > rebalancer. This is intended to be a workaround as the current implementation caused problems when the rebalancing process took a long time. While i also think that having long running custom command is not ideal, this approach seemed to be the right compromise on having some feedback from the process. If two users trigger/start the rebalancing separately, the second will fail with "another balancer is running", as it happens today. - Laszlo ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/57147/#review167117 ----------------------------------------------------------- On Feb. 28, 2017, 6:01 p.m., Laszlo Puskas wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/57147/ > ----------------------------------------------------------- > > (Updated Feb. 28, 2017, 6:01 p.m.) > > > Review request for Ambari, Alejandro Fernandez, Robert Levas, Sandor Magyari, > and Sebastian Toader. > > > Bugs: AMBARI-20175 > https://issues.apache.org/jira/browse/AMBARI-20175 > > > Repository: ambari > > > Description > ------- > > HDFS rebalance operation can last for a long time (hours/days) thus when > triggered from the UI the command may be timed out by the Ambari server. > This behavior may confuse users, making them to trigger the rebalancer again > which will fail with "another balancer is running" error. > > The patch provides support for setting a reasonably long timeout for the > rebalance custo action so that Ambari server doesnt time out the command. > > > Diffs > ----- > > > ambari-server/src/main/java/org/apache/ambari/server/controller/AmbariCustomCommandExecutionHelper.java > 867ebff > > ambari-server/src/main/resources/common-services/HDFS/2.1.0.2.0/metainfo.xml > fd7f2f6 > > > Diff: https://reviews.apache.org/r/57147/diff/1/ > > > Testing > ------- > > Unit tests OK. > Manually tested on local dev-env. > > > Thanks, > > Laszlo Puskas > >