Mageia 4 and rpc.gssd (RPC: AUTH_GSS upcall timed out.)

[feretio]

Hi,

I think this issue may have started in one of the recent updates to Mageia 3 before 4 was officially released but here's the beef:

o NFS mounts via autofs are slow to mount.

o This error appears in the logs when autofs mounts:

Code: Select all

kernel: RPC: AUTH_GSS upcall timed out.
           Please check user daemon is running.

I've searched via Google as well as this forum specifically but all I can find is that I believe it's looking for rpc.gssd to be running. However, though I've found the gss daemon binaries in /usr/sbin (rpc.gssd, rpc.svcgssd, etc. as installed from nfs-utils), I don't seen an obvious way to start the daemon. There's no systemd or init.d service tied to it and I can't find a flag anywhere that enables or disables it from starting (e.g. /etc/sysconfig/nfs).

I did find a setting in /etc/nfsmount.conf to control the RPCGSS security flavors. The default setting is "Sec=sys" but I haven't yet found documentation to describe what that means. I tried setting it to "none" but I still see the aforementioned error when mounting an NFS file system.

I suppose what I'm looking for is decent references on how to properly configure or unconfigure the need for this daemon. Any suggestions are greatly appreciated!

Thanks,

-Jeff

[doktor5000]

Hmmm, seems we're missing something for the package nfs-utils.
Check https://wiki.archlinux.org/index.php/NFS#Client or https://bbs.archlinux.org/viewtopic.php?id=168317
That's you issue right?

[doktor5000]

Actually, the second article is one I found in my earlier research. However, yes, this is what I'm seeing. I didn't see the first article before and didn't think of looking at kernel modules but at least that might be a way to disable it as a workaround for now.

While I haven't found any service that would start the daemon, starting it manually doesn't seem to do anything.

Here's what I see as available services:

Code: Select all

# systemctl list-unit-files | egrep -i 'nfs|gss|nss|rpc'
proc-fs-nfsd.mount                      static
var-lib-nfs-rpc_pipefs.mount            static
nfs-blkmap.service                      disabled
nfs-idmap.service                       static
nfs-lock.service                        static
nfs-mountd.service                      static
nfs-rquotad.service                     static
nfs-secure-server.service               disabled
nfs-secure.service                      disabled
nfs-server.service                      disabled
rpcbind.service                         enabled
rpcbind.socket                          enabled
nfs.target                              disabled
nss-lookup.target                       static
nss-user-lookup.target                  static
rpcbind.target                          static

When running

Code: Select all

sudo /usr/sbin/rpc.gssd -v -v

from the shell, all I see in the log is this:

Code: Select all

Feb 12 13:52:49 myhostname rpc.gssd[23185]: beginning poll

and it doesn't remain running.

[doktor5000]

Please next time use code tags as explained in ftp://ftp5.gwdg.de/pub/linux/mandriva/m ... e_tags.ogv

Maybe you want to try via http://doc.mageia.org/mcc/4/en/content/ ... --nfs.html

[feretio]

Please next time use code tags as explained in

Actually, I did do that in my original post. I just forgot in my reply.

Maybe you want to try via http://doc.mageia.org/mcc/4/en/content/ ... --nfs.html

This link seems to refer to using a GUI to create hard mounts that are stored in fstab (i.e. always mounted). All of my NFS shares are mounted via autofs so they're not mounted all the time.

[wintpe]

since i moved recently from mga2 to 3 i have seen same, takes much longer for my autofs mounts to appear.

not had time yet to diagnose, will check system when i get home and see if im seeing same messages.

Note to myself to try https://bugzilla.redhat.com/show_bug.cgi?id=1001934

regards peter

[doktor5000]

So this should be the fix on the client?

Code: Select all

systemctl enable nfs-secure.service
systemctl start nfs-secure.service

use option -overs=3 when mounting from the client and blacklist kerberos module:

Code: Select all

echo "blacklist rpcsec_gss_krb5" > /etc/modprobe.d/blacklist-nfs-gss-krb5.conf

[wintpe]

only the blacklist is required.

night and day difference.

definitly solved.

regards peter