Date: Wed, 1 Jun 2005 14:03:55 -0700 From: Tim Spencer <tspencer@hungry.com> To: Matthew Jacob <lydianconcepts@gmail.com> Cc: freebsd-scsi@freebsd.org Subject: Re: isp driver + clustered NetApp failover = strangeness Message-ID: <13309938-2F25-4E1B-AF7A-F3BB0036C3E6@hungry.com> In-Reply-To: <7579f7fb05053009267e122cc5@mail.gmail.com> References: <047FCAD3-439C-47EC-A4E4-2253A25CCB39@hungry.com> <7579f7fb05053009267e122cc5@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On May 30, 2005, at 9:26 AM, Matthew Jacob wrote:
> The message you see indicates that the disk, as far as FreeBSD is
> concerned, has gone away. Since we don't really support dynamic
> reattachment, except possible with some cleverness with parts of
> FreeBSD I haven't worked much with (like the new GEOM stuff), it's
> possible you're somewhat toast.
>
Ah well, it was worth a try...
> To do this right involves some more use case scenarios about what can
> occur and deciding what policies to apply. I did this for FreeBSD 4.X
> at a company, but they declined to contribute this work back to
> FreeBSD.
>
That's too bad. Would it help if I were to call them up and ask
nicely? :-)
> A short term fix *might* be (I haven't tested this) to comment out the
> clause around line 2891 in isp_freebsd.c that looks like:
>
> } else {
> xpt_async(AC_LOST_DEVICE, tmppath,
> NULL);
> }
>
Hmm. That seems to be in the 5.x codebase, but not in the 4.x
codebase. I'm actually running 4.11. Do you have a location where I
can make a similar change in the 4.x code? The specific rev of the
driver file that I'm using is:
/* $FreeBSD: src/sys/dev/isp/isp_freebsd.c,v 1.32.2.21 2004/02/02
22:34:48 mjacob Exp $ */
> There's a long history of discussion about this going back some years-
> how much validation do you need to do when something "leaves" and then
> "returns". There are those who believe that WWN validation is enough.
> There are those who believe that WWN validation and checking
> parameters like size is enough. There are those who believe that a
> device that "leaves" and then "returns" has to be treated like a
> complete removal and reattachment event.
Yeah, it would be nice to have the knobs built in for that.
Although multipathing support would probably obviate most of the need
for these features. But that's a whole other story, I'm sure. :-)
Thanks for your help, and have fun!
-tspencer
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?13309938-2F25-4E1B-AF7A-F3BB0036C3E6>