clurmtabd - Cluster NFS Remote Mount Table Daemon
Instances of clurmtabd keep entries relevant to a given mount point in
sync cluster-wide. It does this by polling /var/lib/nfs/rmtab
HOW IT WORKS
A single instance of clurmtabd is spawned by the service manager
(clusvcmgrd(8)) for each mount point in a given user service. Each
instance polls /var/lib/nfs/rmtab every few seconds, looking for
changes. When changes are detected, they are merged in with the
current in-memory view. The entries are filtered according to the
mount point being monitored and are subsequently written out
synchronously to that mount point.
When a user service is relocated (or after a fail-over), clurmtabd
reads the entries on a given mount point and synchronizes them with the
new member’s /var/lib/nfs/rmtab. This is done prior to running the
exportfs(8) command so that clients of the given service do not receive
"Stale NFS file handle" errors.
Clurmtabd is not a part of NFS and can not ensure that "Stale NFS file
handle" errors are never received. It is a best-effort attempt. To
fully prevent "Stale NFS file handle" errors, it is necessary to modify
the Linux kernel as well as nfs-utils.
clurgmgrd(8), rpc.mountd(8), exportfs(8)
Jan 2005 clurmtabd(8)