Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 29 Jan 2004 22:42:35 +0000
From:      "Andrew - Supernews" <andrew@supernews.net>
To:        Pawel Jakub Dawidek <nick@garage.freebsd.pl>
Cc:        freebsd-geom@freebsd.org
Subject:   Re: geom_fox? (multipathing)
Message-ID:  <E1AmKs8-0004xE-Mi@trinity.supernews.net>
In-Reply-To: <20040129213642.GD72053@garage.freebsd.pl> (Pawel Jakub Dawidek's message of "Thu, 29 Jan 2004 22:36:42 %2B0100")
References:  <E1AmImJ-0001In-6t@trinity.supernews.net> <20040129213642.GD72053@garage.freebsd.pl>

next in thread | previous in thread | raw e-mail | index | archive | help
>>>>> "Pawel" == Pawel Jakub Dawidek <nick@garage.freebsd.pl> writes:

 Pawel> Currently multipathing with ISP driver is not possible,
 Pawel> because ISP driver when path is disconnected doesn't cancels
 Pawel> request and doesn't returns an error to upper layers and there
 Pawel> is no chance to detect such situation.

Yes, that was one of the issues I've been looking into.

In the fabric case there are two different failure modes - the
ISP->fabric link can fail (this one currently seems to hang forever
until the link is restored) and the fabric->array link (or the array
processor) can fail, which the ISP driver does detect (but which still
doesn't seem to be handled right). The second of those is actually
more important to me than the first.

If there's anything I can do to help either with the ISP driver, or
testing, let me know...

 Pawel> GEOM_FOX AFAIK is only a proof-of-conecpt.

Yes, I was contemplating making it more than that, but it seems that
this would duplicate part of your work. (You've gone further with it
than I would have needed to go at this stage)

-- 
Andrew, Supernews
http://www.supernews.com



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?E1AmKs8-0004xE-Mi>