[prev in list] [next in list] [prev in thread] [next in thread] 

List:       linux-ha
Subject:    Re: [Linux-HA] LVM usage?
From:       Kirk Ismay <captain () netidea ! com>
Date:       2006-11-30 21:41:26
Message-ID: 456F5006.4070304 () netidea ! com
[Download RAW message or body]

John Hearns wrote:
> A couple of Linux-HA questions tonight.
> Bear with me please!
>
> I'm setting up a two-node active/passive cluster for use in an HPC 
> cluster.
> There is a shared SCSI storage array.
> The primary head node runs as an NFS server, and I'm arranging for the 
> NFS server to fail over. So far so good.
> Using text-mode haresources  (yes, I know), I have:
>
> master  IPaddr2::192.168.1.254/16/eth0:0 \
>         Filesystem::/dev/vg01/users::/users::xfs \
>         Filesystem::/dev/vg02/data1::/data1::xfs \
>         nfsserver \
>         sgemaster
>
I'm working on building a similar configuration, though my goal is to 
deploy Active-Active with each front end getting a separate RAID 
volume.  I'm fairly new at this myself, but I'll share what I've learned 
so far.

In my case I needed to use the fsid= flag in /etc/exports to prevent the 
Stale NFS filehandle problem I was having.
> SO LVM is being used on the master node, and the filesystems are 
> mounting OK.
> Question is - how do I use the LVM script to enable/disable the volumes?
> OR am I misunderstanding?
> This is a SuSE system, and LVM is started on both the master and 
> secondary nodes at boot time, by  /etc/init.d/boot.lvm (??)
> So the same columes are available on each server, ie. I can log in to 
> secondary server and mount the volume 'by hand'
>
> Is it that I should 'switch off' LVM at boot time (/etc/sysconfig flag 
> ??)  The use resource.d/LVM to enable the volumes at heartbeat start ???
I think all that needs to be done is for you to add LVM to your 
haresources like so:

master  IPaddr2::192.168.1.254/16/eth0:0 \
        LVM::/dev/vg01  LVM::/dev/vg02 \
        Filesystem::/dev/vg01/users::/users::xfs \
        Filesystem::/dev/vg02/data1::/data1::xfs \
        nfsserver \
        sgemaster

 From what I can see the heartbeat LVM script marks the volume inactive 
so that while it is visible to lvdisplay -C, you can't actually mount 
it.  When I do an lvscan on my system, the volume assigned to the other 
node shows as inactive:

# lvscan
  ACTIVE            '/dev/storage2/home2' [100.00 GB] inherit
  inactive          '/dev/storage1/home1' [100.00 GB] inherit

Heartbeat will run 'LVM /dev/vg01 stop' on your secondary node on 
startup, marking the volume inactive until a failover event occurs.  
This will keep you from being able to mount the volume by hand.

> Also is the resource.d/SCSI script advisable?
> I am a bit wary of it. Also how do you get the LUNs/IDs of specific 
> disks? Yes, I know it is in /proc, but the recipe? 
In my case, using the LinuxSCSI script wound up causing my kernel to 
mark the volume as dead, and I couldn't remount the volume again without 
a reboot, so I'm not using it at present.

There is an lsscsi utility (in my case available as a Debian package) 
that will show you the LUN information.  The output format is identical 
to what the LinuxSCSI resource script wants:

[0:0:0:0]    disk    SEAGATE  ST336607LC       0007  -
[0:0:1:0]    disk    SEAGATE  ST336607LC       0007  -
[1:0:0:0]    disk    SC U320/ SATA16R          R0.0  -
[1:0:0:1]    disk    SC U320/ SATA16R          R0.0  -
[1:0:0:2]    disk    SC U320/ SATA16R          R0.0  -

 From that, you can add LinuxSCSI::1.0.0.1 or whatever to your 
haresources between the IP and LVM.

Do you have the NFS server working so far? 

Hope that helps.

Sincerely,
Kirk Ismay
System Administrator

-- 
Net Idea
201-625 Front Street Nelson, BC V1L 4B6
P:250-352-3512 | F:250-352-9780 | TF:1-888-352-3512

10 Years of Service Excellence!

Check out our brand new website! www.netidea.com



_______________________________________________
Linux-HA mailing list
Linux-HA@lists.linux-ha.org
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems
[prev in list] [next in list] [prev in thread] [next in thread]