On 07/31/18 08:45, Rick Stevens wrote:
On 07/30/2018 04:45 PM, Ed Greshko wrote:
> On 07/31/18 06:38, Ed Greshko wrote:
>> I'd never seen log entries like the type Bob posted.
>
> Just as a test, I set up a VM as an NFS client with this in the fstab.
>
> ds6:/volume1/misty /home/egreshko/misty nfs4
> rw,soft,intr,fg,comment=systemd.automount 0 0
>
> I then ssh'd into the system as root and....
>
> [root@f28k-b1 ~]# df -T | grep nfs
> [root@f28k-b1 ~]#
>
> Nothing mounted, and
>
> [root@f28k-b1 ~]# journalctl -b 0 | grep misty
> [root@f28k-b1 ~]#
>
> no mention of that filesystem.
>
> Then.....
>
> [root@f28k-b1 ~]# ls /home/egreshko/misty (after a slight delay)
> GPS keys personal Rachel Weather
> ironsocket manuals Pictures-meimei US-Taxes
>
> [root@f28k-b1 ~]# df -T | grep nfs
> ds6:/volume1/misty nfs4 2879621632 932057216 1947462016 33%
/home/egreshko/misty
>
> and
>
> [root@f28k-b1 ~]# journalctl -b 0 | grep misty
> Jul 31 07:40:06
f28k-b1.greshko.com systemd[1]: home-egreshko-misty.automount: Got
> automount request for /home/egreshko/misty, triggered by 1025 (ls)
> Jul 31 07:40:06
f28k-b1.greshko.com systemd[1]: Mounting /home/egreshko/misty...
> Jul 31 07:40:08
f28k-b1.greshko.com systemd[1]: Mounted /home/egreshko/misty.
>
> For completeness the NAS /etc/exports contains....
>
> /volume1/misty
>
*.greshko.com(rw,async,no_wdelay,no_root_squash,insecure_locks,sec=sys,anonuid=1025,anongid=100)
In your fstab entry, "intr" is ignored after kernel 2.6.25 (just a
note--doesn't hurt anything but it's superfluous).
Yeah, creature meet habit. Thanks for pointing that out.
The
comment=systemd.automount
bit in your fstab creates an automount request to systemd which is
activated on the first reference to the mountpoint (which is why it
worked for you after you did the "ls" and not before as shown in your
logs). By the way, the comment bit could also be put in as:
x-systemd.automount
which is a bit clearer to me (it really IS an option, not a comment).
True. And less typing. I think I just mimicked what I saw without doing follow-on
research.
The race condition kicks in when you _don't_ have either of those
in
the fstab entry and you expect the filesystem to mount at boot as it
used to. I mean, automounting is fine and all and this is a bit cleaner
than the old method of automounting, but it's different than how it used
to work and the way systemd and NetworkManager are structured, it's hard
to get the behavior I (for one) expect and it's harder to troubleshoot
(a lot more moving parts working behind the scenes).
I've had no problems with the configuration that I've been using. It has worked
better, for me, than having to make changes to address race conditions.
I just posted the "test" results to show Bob what happens in my environment.
Not
that it should make any difference but my NAS is not Fedora and my NFS client is
getting its address via DHCP.
--
Conjecture is just a conclusion based on incomplete information. It isn't a fact.