On Wed, 16 Mar 2022 10:46:05 -0300, Andreas Hasenack via samba
<samba at lists.samba.org> wrote:
> ctdb from samba 4.15.5
>
> I was looking into ctdb and its NFS integration, after ubuntu/debian
> updated their nfs-utils (nfs server) package to 2.6.x which uses a new
> configuration file (/etc/nfs.conf) instead of /etc/default/nfs-*. It
> mostly works with a few changes, but I have a question about how
> rpc.rquotad is supposed to be run.
>
> In particular, does rpc.rquotad make sense on the NFS server if that
> server is only exporting a glusterfs volume? It's my understanding the
> quota tools don't know about glusterfs, and a gluster volume cannot be
> mounted with the `usrquota,grpquota` mount options like other
> filesystems. Furthermore, at least in debian/ubuntu, the quotarpc
> service (which is what rpc.rquotad goes by) won't start if there are
> no filesystems to be mounted with these quota options.
>
> The situation we get into, then, is that the ctdb script will issue
> "service quotarpc start", but it will exit 0 since there are no
> filesystems with the quota options. Then the event script will
> eventually notice that there is no rpc.rquotad in the rpcinfo output,
> conclude that this service has failed, and try to start again. And it
> loops like that.
>
> If I don't define the nfs_rquotad_service variable, then the ctdb
> script will try to start rpc.rquotad directly. That works, it stays
> running, but it's useless in this scenario, no?
Correct!
The simplest solution would be to remove (or edit) 50.rquotad.check on
your nodes. Hopefully the packager has marked it as a configuration
file, so it would not be replaced on upgrade - if not, that is a
packaging bug.
I've replied in more detail to your similar post to samba-technical, so
we can discuss tweaking the code so this situation doesn't surprise
other people. This reply is for the archives... to help those other
surprised people... :-)
peace & happiness,
martin