The official doc shows in detail how to config and start the hbase cluster, but says little about a shutdown.
I have two tactics to stop the entire cluster:
- stop all the region server first and then stop the master, by script 'hbase-daemon.sh'
- stop master first and then stop all the region server, also by script 'hbase-daemon.sh'
I have tried the first sequence, stop all the region server first and then stop the master. But this could cause a problem: when you close some of all the region server, master would perceive and move the regions of the closed region servers to the rest of the region servers, but the rest of the region servers may not have sufficient memory so that they would crash, rather than exit normally.
The second sequence looks safer than the first, but I haven't given it a try.
Can someone help clarify this?
