> Greetings,
>
> I have been running the xfs filesystem for about 1mo now on a redhat
> 7.0 system. The box was installed using the SGI RedHat install iso, and
> its kernel re-compiled using the CVS tree. Its /usr directory is
> exported to 4 other linux boxes over nfs v2 and v3. We are constanly
> having : nfs, cant get request slot errors between our workstations and
> the server. At times the kernel will dump a stack trace and print out:
> kernel bug 1339! I'm assuming it caused by the md driver, or the nfs
> load, or a combination of both. Any suggestions to a work around? Are
> there any new drivers I wouldn't get from CVS?
> Thanks in advance.
>
> Dave
I have seen this error from my workstation accessing NFS servers which are
not running Linux, it may be a generic problem. This message comes from
this code:
if (clnt->cl_chatty && !(task->tk_flags & RPC_CALL_MAJORSEEN)) {
task->tk_flags |= RPC_CALL_MAJORSEEN;
if (req)
printk(KERN_NOTICE "%s: server %s not responding, still
trying\n",
clnt->cl_protname, clnt->cl_server);
#ifdef RPC_DEBUG
else
printk(KERN_NOTICE "%s: task %d can't get a request
slot\n",
clnt->cl_protname, task->tk_pid);
#endif
}
in an rpc timeout path, so this is a request timeout.
I am not sure if this could be happening because or XFS or not. How recent is
your kernel from cvs, there were changes this week (not complete until
yesterday) which help nfs performance considerably.
Steve
|