Re: Fw: problems with qla2x00
On Tue, Nov 02, 2004 at 07:26:29AM +0100, andreas.sumper@nimbus.at wrote:
> Matt Zagrabelny <mzagrabe@d.umn.edu> wrote on 01/11/2004 11:09:39 PM:
>
Hi,
I'm stepping into this a little late in the game, but since I've been having annoying headaches with qlogic/sun/linux/etc. of late, I thought I would put in my two cents.
> >
> > > > i guess the idea is that the SAN is a single point of failure.
> >
> > >
> > > But aren't single points of failure A Bad Thing?
> >
> > if either A or B fails your system fails is a bad thing. (the system
> > depending on two hard drives)
> >
> > if A fails your system fails is not as bad. (the system depending on one
> > hard drive)
> >
> > >
> > > Does the QLogic SAN driver support redundancy (2 cards, both doing
> > > the same thing, so that if one card dies, the box keeps working)?
> >
> > yes, some cards do (i believe). but this one does not.
It sort of does. The qlogic SAN driver (as of 6.x series) does failover, but only if the port number of the LUNs is the same. It just so happens that this means the card has to share the same loop if in loop mode (ie: a storage box with 4 loops), or do some trickery with the fabric in point to point mode (and I haven't found out what trickery that is, because I don't have a switch or two lying around)
> >
> > > A simple "little" 80GB IDE hard drive doesn't seem that complicated
> > > (unless you have a dozen boxes, which is a good reason for wanting
> > > to boot off the SAN).
> > >
> > > Booting off an IDE drive is about as simple/known/robust as one
> > > could imagine. No need for special drivers, work-arounds, etc.
> > >
> > > Maybe it's just me...
> >
> > the base OS doesnt need much space, so it can live on the SAN with the
> > data. thus the local disk is unnecessary.
>
> The thing for me is, that I have an installed system on an SATA - Disk,
> which works quite well.
> Now I want to use this server as a front end to the Storage...
>
> By now, I found out, that I can not compile qla2x00 (qla2300 for me) on a
> 2.6.8 - smp - Kernel. I had to go back to 2.4.27 - smp, and now I can
> compile the drivers...
Are you using the 8.x driver in 2.6 or something else?
I'm using the qlogic 6.x (and 7.x) series which works ok on RedHat 3, but kind of wigs on vanilla sources from kernel.org, at least when two HBAs are into the same loop to do failover.
I worked around it (when I used to use two loops per server) by using the multipath personality of md in the kernel. It only supports one oopsie, ie: no failback, but it works in case an HBA, cable, controller, switch, etc. wigs.
>
> The problem is, that the debian package only contais source for an old
> driver and the rpm from qlogic did not like my debian :-( But as long as
> there is alien, it does not really hurt ;-)
The rpm from qlogic is still available as a tgz last I downloaded (7.00 or so)
>
>
> Thanks to matt for the hints...
>
>
> Bye,
> Andy
>
> > -matt zagrabelny
> >
> >
> > --
> > To UNSUBSCRIBE, email to debian-user-REQUEST@lists.debian.org
> > with a subject of "unsubscribe". Trouble? Contact
> listmaster@lists.debian.org
> >
> ----------------------------------------------------------------------------
> | Andreas Sumper
> | Administration / Services / IT Security
> | nimbus Development IT Consulting GmbH
> | we unleash the power of domino
> |
> | Annenstrasse 30/1
> | 8020 Graz
> | Tel.: +43 (0) 316 714 255 -> DW 18
> | Fax: +43 (0) 316 714 255 -> DW 4
> | http://www.nimbus.at/
> ----------------------------------------------------------------------------
> | sent through Lotus Notes 6.5.1
> | using wine on a linux box
>
>
--
adam
Reply to: