0

The problem is a little more unusual, i have a glusterfs cluster - suppose with 4 node and some volumes. If one of my node restarts there can be two scenario:

  • The node restarts with the same IP address and all files of previous node, in that case this nodes gets connected to the glusterfs cluster with no other efforts.

  • The nodes restarts with the previous data but this time the IP of the node is changed. So peer status in other nodes shows one disconnected nodes, this is the previous node that was disconnected. But the current node that started is not running. even glusterd failed to run in this node.

My question is how can i connect the new node that already had the data as a replacement for the previous node.

**What I tried so far

  • i deleted the /var/lib/glusterd folder in new node. restarted glusterd. repeer it from another good node. Try replacing the old node with new one, but didn't work.
sadlil
  • 3,077
  • 5
  • 23
  • 36

1 Answers1

0

It's because in other nodes the ip address of the disconnected node is still the same old one, try this:

  • Probe Node2 and Node3 from Node1 as normal ("gluster peer probe node2", "gluster peer probe node3"). After this Node1 will be referred to by it's IP on Node2 and Node3.
  • From one of Node2 or Node3, do a reverse probe on Node1 ("gluster peer probe node1"). This will update the IP to hostname everywhere (in-memory, on-disk and on all nodes).

Ref.: http://lists.gluster.org/pipermail/gluster-users/2015-March/021255.html

Pooria Khodaveissi
  • 395
  • 1
  • 8
  • 16