On 18/1/17 7:51 am, Ed Greshko wrote:
On 01/18/17 04:39, Stephen Morris wrote:
Thanks Ed. I haven't written my own systemd units as I haven't investigated how to do
that, so at the moment I don't have the expertise to do so.
I've written in another response to my original thread that I may have found what the
issue is, but I need to do some more observation, particularly what happens after a cold
start. If the issue has been
fixed I need to determine whether it is standard functionality or a bug.
OK. As I mentioned, I don't have any mnt-nas.mount unit on my system and I can't find any
package in Fedora which would provide that unit. So I just wonder where it comes from.
FWIW, my original entry in my fstab did not have a _netdev parameter specified. I have
since added that and in multiple reboots have had no problems with it getting mounted at
boot time.
From what I have been able to determine from the CIFS mount entry in
fstab and the nfs mount entry in fstab, that mount unit is auto created
by the system from the fstab entries. I have both a mnt-nfs.mount and
mnt-nas.mount unit and from what I have in fstab the information before
the .mount is the mount point. Both those fstab entries are mounting the
same device on mount points /mnt/nfs and /mnt/nas respectively.
What I have found now is that it looks like maybe CIFS has a bug in it
in that it is not honouring the _netdev option. Looking at boot.log,
even though the nfs mount command is in fstab before the cifs mount
command, the system at boot time is trying the mount the cifs entry
first, followed by the nfs entry, and from what I am seeing after a cold
start the cifs mount fails but the nfs mount succeeds.
So what it seems to me at the moment is that CIFS is ignoring the
_netdev parameter and attempting to mount the device immediately whereas
NFS is honouring that parameter. From what happens after a cold start
and a warm start, the network is potentially available much earlier in
the boot process after a warm start, so the CIFS mount works, but after
a cold start where it takes much longer to make the network available
the mount request fails.
Another alternative is that it looks like from boot.log that systemd is
attempting the two mount processes in parallel, and because the nfs
mount is initiated first and it is honouring the _netdev option, it is
locking the device and waiting until the network is available, and then
when the CIFS mount attempts the mount immediately it is locked out by
the nfs mount process, so it times out and hence fails but systemctl
thinks the failure was due instead to the network not being available.
regards,
Steve
_______________________________________________
users mailing list -- users@xxxxxxxxxxxxxxxxxxxxxxx
To unsubscribe send an email to users-leave@xxxxxxxxxxxxxxxxxxxxxxx
