I forgot to mention that, after the test with snapshots, I tried to shutdown the cluster by 'dog cluster shutdown' but only 1 sheep died. The proccess on the other 2 node was frozen. I killed them by kill -9 and restarted the cluster. A killed a node to check if recovery was starting and it started. During recovery I insert back the node (without clearing data first). After sometime, recovery is still running (frozen) on node 1 (I killed node 2). I'm going to format the cluster. |