Red Hat Bugzilla – Bug 1288419
overcloud image contains erroneous rpcbind systemd setup for nfs
Last modified: 2018-02-08 06:05:23 EST
Description of problem:
I have deployed the Overcloud and use NFS for Cinder and Glance.
I found a problem with the NFS mount from the Controller and Compute Nodes.
The 1st NFS mount takes about 1 minute, which is very long and often the start of the Pacemaker FS resource for Glance failed. A subsequent mount will takes less than a second.
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1. Restart a Compute Node
2. Try to do the NFS mount manually:
[root@compute ~]# mount -t nfs nfs-server:/mountpoint /mnt
4. After approx 1 minute the mount command completes successfully
5. The same mount command on the same server, without rebooting the server, will takes less than a second.
1st mount takes ~1min, subsequent mounts <1sec
All mounts, even the first one should be completed in <1sec
See RHEL 7 errata: https://rhn.redhat.com/errata/RHBA-2015-2205.html
and KBase: https://access.redhat.com/solutions/1397363
which describe the problem and also the solution.
The delay also prevents the node from starting up correctly.
Please include this fix in the overcloud image as well.
The referenced bug/solution is shipped as part of RHEL 7.2. The next update of RHEL-OSP will include RHEL 7.2 so this will be fixed then.