Skip site navigation (1)Skip section navigation (2)
Date:      Mon, 14 Aug 2006 08:08:57 +0000
From:      Geoff Buckingham <geoffb@chuggalug.clues.com>
To:        Matthew Jacob <lydianconcepts@gmail.com>
Cc:        freebsd-scsi@freebsd.org
Subject:   Re: isp issues on recent -STABLE
Message-ID:  <20060814080857.GA58022@chuggalug.clues.com>
In-Reply-To: <7579f7fb0608121101g112e006cy1112d282fab753d3@mail.gmail.com>
References:  <44DB8A9C.8090609@centtech.com> <44DC6F9F.4060405@centtech.com> <20060811154348.GA83765@chuggalug.clues.com> <44DCA9F1.1000302@centtech.com> <7579f7fb0608121101g112e006cy1112d282fab753d3@mail.gmail.com>

next in thread | previous in thread | raw e-mail | index | archive | help
> 
> On 8/11/06, Eric Anderson <anderson@centtech.com> wrote:
> >On 08/11/06 10:43, Geoff Buckingham wrote:
> >> On Fri, Aug 11, 2006 at 06:53:03AM -0500, Eric Anderson wrote:
> >>> [..snip..]
> >
> >I have this one host connected to a single QLogic fiber channel switch,
> >which has 5 ACNC fiber channel arrays attached to it, along with a tape
> >robot and tape drives.  Three of the arrays present 3 LUNs each (2TB per
> >LUN), and two of them present 2 LUNs each, 4GB and 10TB - I'm not using
> >these two arrays much yet, and are not really associated with the problems.
> >
> >> man camcontrol, look at the tags section.
> >>
> >
> >Strange that I've never hit this in the past, but now I seem to be
> >hitting it quite often.  The vendor of the arrays says queue depth is
> >256 per LUN, and that coincides with my messages above I believe.
> >
You may want to try reducing the queue size with camcontrol as a potential
work-around to your hangs. I believe the 256 queue size from your logs is
the FreeBSD default, not a value negotiated with your LUN/Array.

A small reduction might help if there was something like an off by one 
error in one of the queue implimentations. A larger reduction might help
if, for example, there were only 256 tags available for the whole array
(giving each of three LUNs around 80 tags each).

The latter scenario may be more likely if you had recently started using
new LUNs on the arrays in question.




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20060814080857.GA58022>