#35 NFS server: down exported chard mount hangs nfs

[David Zafman]

Currently, if a chard mount which is being exported goes down,
and can't failover, nfsd processes will hang. After all the nfsd
processes are hung, the cluster won't respond to any NFS clients
until the down filesystem can be brought back into the cluster.

This is another reason we need a NOHANG type flag which nfsd can
set, so that a down chard filesystem won't hang it. The nfsd
processes would have to get a special error and then drop the
response to the NFS client. This way the nfsd is available to
handle traffic for other available filesystems, and the NFS client
will keep trying.

[Nobody/Anonymous]

Logged In: NO

Also as of SSI-1.1.1 there is a CFS leak/bug that hangs nfsd
instead of returning an error, so eventually all the nfsd
processes hang. The only way out of this particular case is
reboot the initnode.