Having problems with greenplum DB. One one node that sshd went down and we can not connect to that node. Only hard reboot is possible. We try to proper stop greenplum so we want to be able start it again. In GP docs there is no description how to piece-fully stop DB when SSH is not working on one of the nodes.
Asked
Active
Viewed 429 times
2 Answers
0
GPDB relies on SSH for management activities like starting/stopping. So the best option in my opinion would be to stop all the queries running on the cluster, reboot the problematic node, wait for GPDB to failover to mirrors and when the node would be up again execute "gprecoverseg". This way you won't need to shutdown the whole cluster.
In general, you can try something like external web table executing on this specific segment and restarting the sshd itself if your gpadmin user is in sudoers list (external web table executing sudo /etc/init.d/ssh restart on one of the segments running on problematic node)
0x0FFF
- 4,948
- 3
- 20
- 26
0
You can use the gpstop command to stop gp. That's the best way. It will try to stop all the segments. For the broken sshd segment, it will be automatically removed from the cluster by the master. If you configured mirror for the broken server, you won't lose any data.
