Fix #4953
closedceph-deploy: dns mismatches can cause gatherkeys to fail
0%
Description
Our cluster has two networks, a fast network and a control network. The servers names are nodeX for the fast network and nodeX-control for the control network.
Using ceph-deploy to setup a cluster, I issued the following commands.
ceph-deploy new node1-control node2-control node3-control
ceph-deploy mon create
ceph-deploy gatherkeys
the gatherkeys step fails. In looking in /var/run/ceph, the asok files were created with the nodeX name and not the specified nodeX-control.
Upon advice from Sage, I reran this, specifying hosts using the format logicalname:dns name, so the first command turned into this:
ceph-deploy new node1:node1-control node2:node2-control node3:node3-control
This allowed the following steps to complete and the ceph.conf file showed the correct IP addresses (from the control IP block).
It seems like the ceph-deploy process could error out if there is confusion when resolving the hostnames into IP addresses, such as in our case where the name and IP address did not match.