[linux-lvm] alternate link / multi-pathing

[linux-lvm] alternate link / multi-pathing

[C'est Pierre]

Hello,

How can I setup multi-pathing so that I can create a vg with a
physical device over two different fibre-channel links, so if one link
fails, LVM uses the other?

I am using the term alternate link but its origins are in hp-ux.

If you could point me to some documentation on it, I would appreciate it!

Thanks,
Pierre

[linux-lvm] Re: alternate link / multi-pathing

[C'est Pierre]

Nevermind about my previous question! I just made my way into this
page: http://www.redhat.com/docs/manuals/enterprise/RHEL-4-Manual/s390-multi-install-guide/s1-s390info-raid.html
and have setup the multipathing nicely.

However, I went into the second node (this is 2 nodes failover cluster
environment) and I noticed that the vg and the md0 appeared there
there automatically! is this normal? is there anything I should do?

Thanks,
Pierre



On 5/28/06, C'est Pierre <cestpierre@gmail.com> wrote:
> Hello,
>
> How can I setup multi-pathing so that I can create a vg with a
> physical device over two different fibre-channel links, so if one link
> fails, LVM uses the other?
>
> I am using the term alternate link but its origins are in hp-ux.
>
> If you could point me to some documentation on it, I would appreciate it!
>
> Thanks,
> Pierre
>

[linux-lvm] Re: alternate link / multi-pathing

[C'est Pierre]

Again, this time I am looking for some clues. I was trying to make a
test, so I created a logical volume called 'lv01', mounted it under a
directory and created a file on it, but somehow, when I am trying to
unmount it, the umount command stops responding and all I get is these
messages when I run dmesg:

multipath: sdb: redirecting sector 4480 to another IO path
multipath: only one IO path left and IO error.
multipath: sdb1: rescheduling sector 4489
multipath: sdb: redirecting sector 4480 to another IO path
multipath: only one IO path left and IO error.
multipath: sdb1: rescheduling sector 4489
multipath: sdb: redirecting sector 4480 to another IO path
multipath: only one IO path left and IO error.
multipath: sdb1: rescheduling sector 4489
multipath: sdb: redirecting sector 4480 to another IO path
multipath: only one IO path left and IO error.


Any hints? the load average is going pretty crazy too: load average:
7.22, 5.00, 2.24

Thanks,
Pierre


On 5/28/06, C'est Pierre <cestpierre@gmail.com> wrote:
> Hello,
>
> How can I setup multi-pathing so that I can create a vg with a
> physical device over two different fibre-channel links, so if one link
> fails, LVM uses the other?
>
> I am using the term alternate link but its origins are in hp-ux.
>
> If you could point me to some documentation on it, I would appreciate it!
>
> Thanks,
> Pierre
>

Re: [linux-lvm] Re: alternate link / multi-pathing

[Luca Berra]

On Sun, May 28, 2006 at 02:50:24AM +0100, C'est Pierre wrote:
>Nevermind about my previous question! I just made my way into this
>page: 
>http://www.redhat.com/docs/manuals/enterprise/RHEL-4-Manual/s390-multi-install-guide/s1-s390info-raid.html
>and have setup the multipathing nicely.

actually the device-mapper-multipath would a better approach than
md-multipath

>However, I went into the second node (this is 2 nodes failover cluster
>environment) and I noticed that the vg and the md0 appeared there
>there automatically! is this normal? is there anything I should do?

it is not normal
it should not happen

if you insist in using md multipath, ensure raid autodetect won't pick
it up.
i.e. disable damned autodetect on kernel command line or
change the partition type to something different than 'Linux Raid'

then you'll have to ensure that your distribution  init scripts do not
automatically start it.
best way is omitting this set entries from /etc/mdadm.conf
and putting it somewhere else, then invoking
"mdadm -Asc /some/where/else/mdadm.conf" in your cluster script

if you use device-mapper-multipath (which might also solve the issue in
your next post :)
there is no problem if multipath finds the devices on the failover node,
but you must prevent lvm from activating the vg

to do this either add cluster locking or use tags, as explained in the 
doc/tagging.txt files in the lvm distribution

Regards,
L.


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

Re: [linux-lvm] Re: alternate link / multi-pathing

[C'est Pierre]

Hello Luca,

Thanks for pointing me that out, too bad it's a bit late, since the
server is now crashed and only tomorrow I will be able to reset it.
Meanwhile, I've got a few questions for you:

1. How do I disable raid autodetect ?
2. Where can I read about device-mapper-multipath ?
3. To deactivate the vg, I suppose I should use vgchange -a n, correct
me if I am wrong

Thank you for your reply!
Gon�alo


On 5/28/06, Luca Berra <bluca@comedia.it> wrote:
> On Sun, May 28, 2006 at 02:50:24AM +0100, C'est Pierre wrote:
> >Nevermind about my previous question! I just made my way into this
> >page:
> >http://www.redhat.com/docs/manuals/enterprise/RHEL-4-Manual/s390-multi-install-guide/s1-s390info-raid.html
> >and have setup the multipathing nicely.
>
> actually the device-mapper-multipath would a better approach than
> md-multipath
>
> >However, I went into the second node (this is 2 nodes failover cluster
> >environment) and I noticed that the vg and the md0 appeared there
> >there automatically! is this normal? is there anything I should do?
>
> it is not normal
> it should not happen
>
> if you insist in using md multipath, ensure raid autodetect won't pick
> it up.
> i.e. disable damned autodetect on kernel command line or
> change the partition type to something different than 'Linux Raid'
>
> then you'll have to ensure that your distribution  init scripts do not
> automatically start it.
> best way is omitting this set entries from /etc/mdadm.conf
> and putting it somewhere else, then invoking
> "mdadm -Asc /some/where/else/mdadm.conf" in your cluster script
>
> if you use device-mapper-multipath (which might also solve the issue in
> your next post :)
> there is no problem if multipath finds the devices on the failover node,
> but you must prevent lvm from activating the vg
>
> to do this either add cluster locking or use tags, as explained in the
> doc/tagging.txt files in the lvm distribution
>
> Regards,
> L.
>
>
> --
> Luca Berra -- bluca@comedia.it
>         Communication Media & Services S.r.l.
>  /"\
>  \ /     ASCII RIBBON CAMPAIGN
>   X        AGAINST HTML MAIL
>  / \
>
> _______________________________________________
> linux-lvm mailing list
> linux-lvm@redhat.com
> https://www.redhat.com/mailman/listinfo/linux-lvm
> read the LVM HOW-TO at http://tldp.org/HOWTO/LVM-HOWTO/
>

Re: [linux-lvm] Re: alternate link / multi-pathing

[Luca Berra]

On Sun, May 28, 2006 at 03:35:38PM +0100, C'est Pierre wrote:
>Hello Luca,
>
>Thanks for pointing me that out, too bad it's a bit late, since the
>server is now crashed and only tomorrow I will be able to reset it.
>Meanwhile, I've got a few questions for you:
>
>1. How do I disable raid autodetect ?
/usr/src/linux/Documentation/md.txt

>2. Where can I read about device-mapper-multipath ?
http://christophe.varoqui.free.fr/
Redhat has a device-mapper-multipath rpm ensure you are running RHEL4-U3

>3. To deactivate the vg, I suppose I should use vgchange -a n, correct
>me if I am wrong
yes, but i do not get the connection with your first question.

>Thank you for your reply!
you are welcome,
but please do not top post

L.


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