The main problem is that the credentials used for the SST are incorrect. It can't access the data so it can't transfer the state to other nodes.
For the point about needing to start the first node with the cluster address of gcomm://
, that is needed when there are no running nodes in the cluster to connect to. I'm not sure about the init script on Ubuntu, but I know that the init script on RHEL and CentOS has a bootstrap
command that forces gcomm://
as the address for starting a new cluster. Check to see if yours has that and if so, keep the address to connect to a running cluster in the config and use the bootstrap
command in the init script to bootstrap the cluster. If yours doesn't have the bootstrap
command, you should be able to add the following into the init script to add it:
'bootstrap')
# Bootstrap the cluster, start the first node
# that initiate the cluster
echo $echo_n "Bootstrapping the cluster"
$0 start $other_args --wsrep-new-cluster
;;
That's from my RHEL init script and it might need to be modified slightly, but that's at least generally what it needs.