Sorry, I should have mentioned before -- we are using CDH 4.2, which does not package the graceful_stop script. Do you happen to know if there's a way to do this through the CDH manager? Perhaps the "decommission" action does something similar? My impression is that "decommission" is more heavy-handed, but if that's the most convenient route, that'll work for us.
Thanks, - Dan On Apr 25, 2013, at 11:30 AM, Jean-Marc Spaggiari wrote: > Hi Dan, > > You might want to take a look at bin/graceful_stop.sh . It will move > all the regions hosted by your RS to other RS before stopping it > gracefuly. After the maintenance, simply start the RS/DN back and it > will be added back to the cluster. Loadbalancer will then assign some > regions back to him. You will loose some data locality for the regions > wich are going to be moved. > > JM > > 2013/4/25 Dan Crosta <[email protected]>: >> We have to perform maintenance on one of our HDFS DataNode/HBase >> Regionserver machines for a few hours. What are the right steps to take >> before doing the maintenance in order to ensure limited impact to the >> cluster and (thrift) clients of the cluster, both for HDFS and HBase? >> >> After the maintenance, are there any special steps required to add the node >> back to the cluster, or can we simply restart the services and HDFS/HBase >> take care of the rest? >> >> Thanks, >> - Dan
