Hi,
On Fri, Feb 01, 2013 at 08:24:46AM +0100, Ulrich Windl wrote:
> Hi!
>
> While trying to develop an improved exportfs RA that can export one
> filesystem to a list of names (instead of just one name), I found an error:
>
> exportfs is returning exit code 0 even if the filesystem could not be
> exported, like in
>
> > h02:~ # exportfs h012:/mnt; echo $? # h012 does not exist
> > exportfs: Failed to resolve h012
> > 0
>
> Accordingly the start operation for the exportfs alway reports success, even
> if the operation failed! That's due to
>
> [...]
> > ocf_run exportfs -v ${OPTIONS}
> > ${OCF_RESKEY_clientspec}:${OCF_RESKEY_directory} || exit $OCF_ERR_GENERIC
> >
> >
> > # Restore the rmtab to ensure smooth NFS-over-TCP failover
> > restore_rmtab
> >
> > ocf_log info "File system exported"
> > return $OCF_SUCCESS
> [...]
>
> However the monitor operation does the correct thing, thus reporting an error
> (rc==7).
We could insert the monitor operation after the call to exportfs.
Would there be any timing issues? I guess not.
Thanks,
Dejan
> This combination leads to the message of ocf-tester:
> * rc=1: Monitoring an active resource should return 0
>
> (The tester thinks the resource was started when it was not)
>
> I had reported this for SLES10 SP2 to support in November, but after a long
> time of waiting and repeating the same facts over and over, support told me
> it's "working as designed"; lvscan and lvcreate would be similar.
>
> A quick test showed that this is actually not true vor lvcreate:
> # lvcreate -n bar -L40G foobar; echo $?
> Volume group "foobar" not found
> 5
> # lvcreate -n foo -L1T sys; echo $?
> Volume group "sys" has insufficient free space (1855 extents): 32768
> required.
> 5
>
> I got the impression that my support is just unwilling to fix the rather
> trivial problem. My exportfs comes from nfs-kernel-server-1.2.3-18.23.1.
>
> Getting the exportfs sources, it took me two minutes to find out that
> exportfs uses variable export_errno to define the exit code, and that
> variable is nowhere set. The worker routine exportfs() does return nothing at
> all. What a design!
>
> Here's an example from an old HP-UX system that didn't get any updates for
> three years:
>
> # exportfs nohost:/mnt; echo $?
> exportfs: no entry for nohost:/mnt in /etc/dfs/dfstab
> 1
>
> Here it works!
>
> I wonder why some Linux people are so ignorant occasionally...
>
> Regards,
> Ulrich
>
>
> _______________________________________________
> Linux-HA mailing list
> [email protected]
> http://lists.linux-ha.org/mailman/listinfo/linux-ha
> See also: http://linux-ha.org/ReportingProblems
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems
