Fw: Multipath and SAN Fabric failover

Fw: Multipath and SAN Fabric failover

[John Little]

Bernd/All,

Some additional infos:

I unplugged the fc cable from fabric a and rebooted.  When checking
dmesg I noticed that there were not any dm-emc parameters loaded like
there were in my previous email below when rebooting with both cables
connected.

xen2:~ # dmesg |grep -i device-mapper
device-mapper: 4.5.0-ioctl (2005-10-04) initialised: dm-devel@redhat.com
device-mapper: dm-multipath version 1.0.4 loaded
device-mapper: dm-round-robin version 1.0.0 loaded
device-mapper: dm-multipath: error getting device
device-mapper: error adding target to table
xen2:~ #

Am I missing some setup somewhere for a dual port hba?

Thanks,
John

Begin forwarded message:

Date: Fri, 8 Sep 2006 11:32:32 -0400
From: John Little <jslittl@hendricks.org>
To: dm-devel@redhat.com
Subject: Re: [dm-devel] Multipath and SAN Fabric failover


Hi Bernd

I commented out everything including the blacklist in the
multipath.conf as you suggested.  When I unplug the fc cable from
fabric a the filesystem goes into read only mode.  multipath -v2 -d
shows the same errors previously noted.  After rebooting I unplugged
the fc cable from fabric b.  The file system properly stays in
read\write mode.

Just to make sure that I've covered everything I partitioned the disk
as /dev/sdc and then used lvm2 command:
pvcreate /dev/disk/by-name/capsprod-part1 to initialize and use the
partition with lvm.

I found the following error in dmesg:

device-mapper: 4.5.0-ioctl (2005-10-04) initialised: dm-devel@redhat.com
dm-netlink version 0.0.2 loaded
device-mapper: dm-multipath version 1.0.4 loaded
device-mapper: dm-round-robin version 1.0.0 loaded
device-mapper: dm-multipath: error getting device<====error
device-mapper: error adding target to table<====error
device-mapper: dm-emc version 0.0.3 loaded
device-mapper: dm-emc: long trespass command will be send
device-mapper: dm-emc: honor reservation bit will not be set (default)
device-mapper: dm-emc: long trespass command will be send
device-mapper: dm-emc: emc_pg_init: sending switch-over command
device-mapper: dm-emc: honor reservation bit will not be set (default)
device-mapper: dm-emc: emc_pg_init: sending switch-over command

Can you give me an idea of what else to check?  I'm nowhere near an
expert on this subject as just about everything I've learned has been
from reading this list.  

A couple  of questions come to mind:
Is the lun not properly trespassing  from fabric a to fabric b?
Where does the directive [features=1 queue_if_no_path] come from when
running multipath -v2 -d?  Could this be the problem?

Thanks,

John


>Heya,


>I think you should just try an empty multipath.conf, only put the
>blacklist in. The EMC CX needs pp_emc as prio callout for example. The
>defaults within >multipath should just work. Also make sure your
>failovermode and arraycommpath are both set to 1. A few hints are
>mentioned on
>>http://christophe.varoqui.free.fr/wiki/wakka.php?wiki=TestedEnvironments

>Please fill in if you find something that's missing there.

>Best regards,

>Bernd Zeimetz


>>John Little wrote:

>>    Hi all,

>>    I am trying to setup our server so that we have fabric failover
>> from
 >>   the san.  However my file sytem comes up read-only when I pull a
cable >>  for testing.

>>    I am using the following:
>>    SLES 10
>>    multipath-tools-0.4.6-25.8
>>    q-logic 2342 dual port hba
>>    emc cx600 san
>>    (4) McData DS-24 switches (2 for fabric a and 2 for fabric b)

Re: Fw: Multipath and SAN Fabric failover

[Bernd Zeimetz]

Heya,
> Just to make sure that I've covered everything I partitioned the disk
> as /dev/sdc and then used lvm2 command:
>   
sorry, I've missed that before. /dev/sdc is only _one_ of your paths to 
a LUN. multipath will take the right 2 of those sd* to create one dm 
device file. Have a look if there's somethign in /dev/mapper, also check 
your udev rules, you need this line in there:

KERNEL=="dm-[0-9]*", PROGRAM="/sbin/devmap_name %M %m", NAME="%k", 
SYMLINK="%c"

most distros have a line like
#KERNEL=="dm-[0-9]*",           NAME=""
in there, you need to comment it out if you add the lien above.


Hope that helped,

best regards

Bernd Zeimetz

Re: Fw: Multipath and SAN Fabric failover

[John Little]

dm-devel-bounces@redhat.com wrote on 09/10/2006 03:23:55 PM:

> Heya,
> > Just to make sure that I've covered everything I partitioned the
> > disk as /dev/sdc and then used lvm2 command:
> >   
> sorry, I've missed that before. /dev/sdc is only _one_ of your paths
> to a LUN. multipath will take the right 2 of those sd* to create one
> dm device file. Have a look if there's somethign in /dev/mapper, also
> check your udev rules, you need this line in there:
> 
> KERNEL=="dm-[0-9]*", PROGRAM="/sbin/devmap_name %M %m", NAME="%k", 
> SYMLINK="%c"
> 
> most distros have a line like
> #KERNEL=="dm-[0-9]*",           NAME=""
> in there, you need to comment it out if you add the lien above.
> 
> 
> Hope that helped,
> 
> best regards
> 
> Bernd Zeimetz

Bernd,

I haven't had a chance to work on this for a couple of days but I did
get do some reading.  I'm wondering if I'm using the correct driver for
this.  I have been trying to use the QLogic driver shipped with SLES.
Should I be using the non-EMC driver from the Qlogic website?

Regards,
John

Re: Fw: Multipath and SAN Fabric failover

[Bernd Zeimetz]

hi,

> I haven't had a chance to work on this for a couple of days but I did
> get do some reading.  I'm wondering if I'm using the correct driver for
> this.  I have been trying to use the QLogic driver shipped with SLES.
they should work fine (I hope...). But did you check what I've written 
in my last mail?


Best regards,

Bernd