SuSE Linux cluster using MC/ServiceGuard --- shared storage

SuSE Linux cluster using MC/ServiceGuard --- shared storage

[Biju A]

Hi,

In a Linux cluster setup, I have some issues.

I am trying to use 'md' driver in Linux for multipath
to shared storage with Enterprise Virtual Array 3000
(EVA 3000)

The O/S is SuSE Linux ES 8.0 which runs on both the
cluster nodes (HP ProLiant ML330 machines). I use
Qlogic HBAs (of different versions) both supporting
failover. 

Do I still need to use the Driver Patch for EVA
Controller Failover? 

I can add the Virtual Disks from the EVA console and
this does not need additional Secure Path installation
to get detected.

EVA offers different RAID functions such as 0,1 and 5
and I guess I use the software RAID on Linux
(implemented thru md drivers) to provide multipath
support. Is there any need for implementing a software
RAID on top of an EVA which already has hardware RAID
functionality (say for example, if I want mirroring
instead of multipath)

The MC/SG spec says that 'md' driver should not be
active at the same time on both the cluster nodes. So,
I am not using the 'fd' type of partition which starts
the arrays at boot time. I just use the default 83
type.

Is there any specific need to update to 2.6 kernel in
order to support LVM 2?

The kernel version of the O/S in the cluster nodes is
2.4.19 which contains lvm 1.0.5-51. It does not
support 'pvremove' and probably the tool 'mkraid' does
have some bugs. No idea if there are fixes for the
same. Is it better to use 'mdadm'? 

I have tried to do the same sequence of steps listed
in "Managing Service Guard for Linux" to create the
logical volume infrastructure. But the first disk
partition grouped using md does not become part of the
array --- after rebooting the nodes. After rebooting,
the second node should get the configuration from the
primary node replicated but this does not happen for
the first partition. Only the second partition seems
to be shown active.  

Any hints/suggestions would be appreciated.

Regards,
Biju A




		
__________________________________
Do you Yahoo!?
Take Yahoo! Mail with you! Get it on your mobile phone.
http://mobile.yahoo.com/maildemo 

Re: SuSE Linux cluster using MC/ServiceGuard --- shared storage

[Luca Berra]

On Mon, Sep 20, 2004 at 06:28:46AM -0700, Biju A wrote:
>Hi,
>
>In a Linux cluster setup, I have some issues.
>
>I am trying to use 'md' driver in Linux for multipath
>to shared storage with Enterprise Virtual Array 3000
>(EVA 3000)

i believe someone from HP already stated in this list that md driver
(multipath) does not work with EVA, i guess do to issues with
preferred path.

>Do I still need to use the Driver Patch for EVA
>Controller Failover? 
i believe using the qlogic driver failover is a viable solutio, another
is purchasing securepath from HP. but i believe there were issues using
securepath and lvm together.

>The MC/SG spec says that 'md' driver should not be
>active at the same time on both the cluster nodes. So,
>I am not using the 'fd' type of partition which starts
>the arrays at boot time. I just use the default 83
>type.
>
>Is there any specific need to update to 2.6 kernel in
>order to support LVM 2?

i don't know if that is supported by either hp or suse.

>The kernel version of the O/S in the cluster nodes is
>2.4.19 which contains lvm 1.0.5-51. It does not
>support 'pvremove' and probably the tool 'mkraid' does
>have some bugs. No idea if there are fixes for the
>same. Is it better to use 'mdadm'? 

i do personallu prefer mdadm to mkraid.

>Any hints/suggestions would be appreciated.
also try asking in the hp itrc forums.

L.

-- 
Luca Berra -- bluca@comedia.it
        Communication Media & Services S.r.l.
 /"\
 \ /     ASCII RIBBON CAMPAIGN
  X        AGAINST HTML MAIL
 / \

Re: SuSE Linux cluster using MC/ServiceGuard --- shared storage

[Lars Marowsky-Bree]

On 2004-09-20T15:55:51,
   Luca Berra <bluca@comedia.it> said:

> >Do I still need to use the Driver Patch for EVA
> >Controller Failover? 
> i believe using the qlogic driver failover is a viable solutio,

The DM multipath in 2.6 should handle the EVA just fine.

The qlogic internal failover is being depreciated and AFAIK isn't
included in 2.6 mainline even...

> another is purchasing securepath from HP. but i believe there were
> issues using securepath and lvm together.

Proprietary multipathing solutions are a thing of the past on Linux, one
hopes.


Sincerely,
    Lars Marowsky-Brée <lmb@suse.de>

-- 
High Availability & Clustering
SUSE Labs, Research and Development
SUSE LINUX AG - A Novell company

-
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: SuSE Linux cluster using MC/ServiceGuard --- shared storage

[Luca Berra]

On Mon, Sep 20, 2004 at 08:33:18PM +0200, Lars Marowsky-Bree wrote:
>On 2004-09-20T15:55:51,
>   Luca Berra <bluca@comedia.it> said:
>
>> >Do I still need to use the Driver Patch for EVA
>> >Controller Failover? 
>> i believe using the qlogic driver failover is a viable solutio,
>
>The DM multipath in 2.6 should handle the EVA just fine.

you mean multipath-tools (http://christophe.varoqui.free.fr/multipath.html)?
i suppose so, using the "group_by_tur" spreading policy. fortunately i
don't have any spare EVA to test, plenty of XP tough :)

>The qlogic internal failover is being depreciated and AFAIK isn't
>included in 2.6 mainline even...

excuse me, are you shipping (or at least supporting) a 2.6 kernel on
your SLES 8.0 product?

>> another is purchasing securepath from HP. but i believe there were
>> issues using securepath and lvm together.
>
>Proprietary multipathing solutions are a thing of the past on Linux, one
>hopes.
i hope so, but for this statement to become true we need some other
vendor to ship a 2.6 based 'enterprise level' distribution with proper
device-mapper implementation, and application vendors (that includes
HP and IBM) to adapt.
Anyway, i do not really believe securepath to be a viable solution(*), i
was enumerating possibilities.

(*) and hp suggests using qlogic 7.00.3 driver with internal failover
for use with their Serviceguard product.

-- 
Luca Berra -- bluca@comedia.it
        Communication Media & Services S.r.l.
 /"\
 \ /     ASCII RIBBON CAMPAIGN
  X        AGAINST HTML MAIL
 / \

Re: SuSE Linux cluster using MC/ServiceGuard --- shared storage

[Lars Marowsky-Bree]

On 2004-09-21T08:34:12,
   Luca Berra <bluca@comedia.it> said:

> >>>Do I still need to use the Driver Patch for EVA
> >>>Controller Failover? 
> >>i believe using the qlogic driver failover is a viable solutio,
> >The DM multipath in 2.6 should handle the EVA just fine.
> you mean multipath-tools (http://christophe.varoqui.free.fr/multipath.html)?
> i suppose so, using the "group_by_tur" spreading policy. fortunately i
> don't have any spare EVA to test, plenty of XP tough :)

You could always ask HP to give you one as a loaner while you make the
code work ;-)

> >The qlogic internal failover is being depreciated and AFAIK isn't
> >included in 2.6 mainline even...
> excuse me, are you shipping (or at least supporting) a 2.6 kernel on
> your SLES 8.0 product?

Not sure how it relates to this comment, but no, SLES8 is 2.4 only.
However, we do support 2.6 on our SLES9 platform, which has been out
since something like 1-2 months. (In R&D, project release dates become a
blur.)


Sincerely,
    Lars Marowsky-Brée <lmb@suse.de>

-- 
High Availability & Clustering
SUSE Labs, Research and Development
SUSE LINUX AG - A Novell company

-
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

RE: SuSE Linux cluster using MC/ServiceGuard --- shared storage

[Stern, Rick (Serviceguard Linux)]

It is my understanding that MD/DM does not have the specific commands to
support controller failover in the EVA.  

-----Original Message-----
From: linux-raid-owner@vger.kernel.org
[mailto:linux-raid-owner@vger.kernel.org] On Behalf Of Luca Berra
Sent: Monday, September 20, 2004 11:34 PM
To: linux-raid@vger.kernel.org
Subject: Re: SuSE Linux cluster using MC/ServiceGuard --- shared storage

On Mon, Sep 20, 2004 at 08:33:18PM +0200, Lars Marowsky-Bree wrote:
>On 2004-09-20T15:55:51,
>   Luca Berra <bluca@comedia.it> said:
>
>> >Do I still need to use the Driver Patch for EVA
>> >Controller Failover? 
>> i believe using the qlogic driver failover is a viable solutio,
>
>The DM multipath in 2.6 should handle the EVA just fine.

you mean multipath-tools
(http://christophe.varoqui.free.fr/multipath.html)?
i suppose so, using the "group_by_tur" spreading policy. fortunately i
don't have any spare EVA to test, plenty of XP tough :)

>The qlogic internal failover is being depreciated and AFAIK isn't
>included in 2.6 mainline even...

excuse me, are you shipping (or at least supporting) a 2.6 kernel on
your SLES 8.0 product?

>> another is purchasing securepath from HP. but i believe there were
>> issues using securepath and lvm together.
>
>Proprietary multipathing solutions are a thing of the past on Linux,
one
>hopes.
i hope so, but for this statement to become true we need some other
vendor to ship a 2.6 based 'enterprise level' distribution with proper
device-mapper implementation, and application vendors (that includes
HP and IBM) to adapt.
Anyway, i do not really believe securepath to be a viable solution(*), i
was enumerating possibilities.

(*) and hp suggests using qlogic 7.00.3 driver with internal failover
for use with their Serviceguard product.

-- 
Luca Berra -- bluca@comedia.it
        Communication Media & Services S.r.l.
 /"\
 \ /     ASCII RIBBON CAMPAIGN
  X        AGAINST HTML MAIL
 / \
-
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: SuSE Linux cluster using MC/ServiceGuard --- shared storage

[Lars Marowsky-Bree]

On 2004-09-21T06:40:43,
   "Stern, Rick (Serviceguard Linux)" <rick.stern@hp.com> said:

> It is my understanding that MD/DM does not have the specific commands to
> support controller failover in the EVA.  

I thought for the EVA none was needed, just that the overhead associated
with switching was too high and thus one should only stick to the
preferred path until it failed? (ie, no load-balancing at all.)


Sincerely,
    Lars Marowsky-Brée <lmb@suse.de>

-- 
High Availability & Clustering
SUSE Labs, Research and Development
SUSE LINUX AG - A Novell company

-
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html