Verry serious problem with ZFS & 12.0

[Julien Cigar]

On Mon, Sep 09, 2019 at 01:45:52PM +0200, Albert Shih wrote:
> Le 29/08/2019 à 10:37:28+0200, Julien Cigar a écrit
> > On Thu, Aug 29, 2019 at 12:45:47AM +0200, Albert Shih wrote:
> > > Hi
> > >
> > > After update 4 servers from 11.2 to 12.0 without any problem, wait few
> > > weeks to see if everything work well, and it did. I just upgrade my mail
> > > server.
> > >
> > > During the upgrade I also upgrade all firmware for the hardware.
> > >
> > > And now I got a very serious issue with my server.
> > >
> > > Configuration :
> > >
> > >   Dell PowerEdge R740Xd with H730P, 192 Go Ram, 2 SAS mechanical disk for the system,
> > >   2 SSD (in a zfs pool) for the mail index (cyrus), and 28 mechanical disk
> > >   (in a second zfs pool) for the mailbox.
> > >
> > > The problem:
> > >
> > >   After running few days the zfs pool with the 2 SSD are not responding.
> > >
> > >   The system are perfectly working.
> > >
> > >   The second zpool (mechanical disk) are perfectly working.
> > >
> > >   I got zero log, zero message in the console or in dmesg.
> > >
> > >   The arc_size are correct, it's around 70-75 %.
> > >
> > >   The moment the zfs pool become not responding are random, not related to
> > >   any activity (human or cron).
> > >
> > >   The only option I pass for the kernel related to ZFS are vfs.zfs.min_auto_ashift=12 and
> > >   vfs.zfs.prefetch_disable=1. Without the second one the system no
> > >   responding (under 11.2) when the server send (through zfs send) the data to another
> > >   server.
> > >
> > >   After the first problem I make a zfs upgrade, thinking maybe that's the
> > >   problem so I'm not sure I can downgrade to 11.2 (and 11.2 are EOL)
> > >
> > > In your opinion :
> > >
> > >   1/ What should I do to try to find the problem ?
> > >
> > >   2/ Do you think that's a hardware/firmware problem or FreeBSD problem,
> > >   the point is the second zpool are working perfectly so I'm thinking at
> > >   some firmware/hardware/compatibility problem.
> > >
> > >
> > > Regards.
> >
> > looks like PR 236480
> >
> > see https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=236480
> >
> 
> So I can confirm, with this patch the server work fine without any hang or
> crash.

Cool, it would be worth to add a comment to the PR (problem was observed
with PostgreSQL only until now)

> 
> Thanks folks.
> 
> Regards
> 
> --
> Albert SHIH
> Observatoire de Paris
> xmpp: jas at obspm.fr
> Heure local/Local time:
> Mon 09 Sep 2019 01:44:42 PM CEST

-- 
Julien Cigar
Belgian Biodiversity Platform (http://www.biodiversity.be)
PGP fingerprint: EEF9 F697 4B68 D275 7B11  6A25 B2BB 3710 A204 23C0
No trees were killed in the creation of this message.
However, many electrons were terribly inconvenienced.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: not available
URL: <http://lists.freebsd.org/pipermail/freebsd-questions/attachments/20190909/5deadf91/attachment.sig>