[ClusterLabs] Antw: CentOS 7 - Pacemaker - Problem with nfs-server and system

[Dennis Jacobfeuerborn]

On 08/12/2015 10:35 AM, Stefan Bauer wrote:
> Hi,
> 
> 
> 
> thank you for your reply. It seems to be a problem with the systemd unit files for nfs-server - specifically a timing issue.
> 
> 
> 
> [root at centos7-n1 ~]# systemctl list-unit-files --type=service | grep rpcbind
> rpcbind.service                             static
> 
> 
> 
> rpcbind is set to static - should be started on demand by other units.
> 
> 
> 
> Invoking systemctl start nfs-server is pulling in rpcbind and nfs-lock
> 
> 
> 
> rpcbind is started - but nfs-lock is maybe trying to start too early:
> 
> 
> 
> Invoking manually systemctl start rpcbind and then systemctl start nfs-lock works within a second.
> 
> Invoking manually systemctl start rpcbind and then sytemctl start nfs-server works within a few seconds as well.
> 
> 
> 
> Invoking manually systemctl start nfs-server is only working randomly due to some timing issues.
> 
> 
> 
> My current workaround is to also start rpcbind by the cluster - just before nfsserver.
> 
> 
> 
> I also tried /usr/lib/ocf/resource.d/heartbeat/nfsserver - it is capable of handling systemd systems but start nfs-lock and nfs-server manually - hence hit the same problem in my case.
> 

You are getting bitten by this RHEL/CentOS7 bug:
https://bugzilla.redhat.com/show_bug.cgi?id=1171603

You need to copy the following files from /usr/lib/systemd/system/ to
/etc/systemd/system/:
nfs-server.service
nfs-lock.service
rpc-statd.service

and then replace "rpcbind.target" with "rpcbind.service" in all three
files. After that you will no longer run into the 1 minuten timeout and
the nfs-server resource script should work better.

Regards,
  Dennis