SUMMARY: Veritas non-rootdg vol not seen after reboot

From: Rasana Atreya (rasana_atreya@hotmail.com)
Date: Mon Sep 20 1999 - 12:24:56 CDT


Hi all,

Here's my (almost) summary on Veritas's volumes not being seen. My orginal
post is marked by ">".

>I used Veritas Volume Manager 3.0.1 to setup a rootdg group to mirror >the
>rootdisk (internal disks), and non-rootdg group with RAID0+1 on a A5000
>array. The box is UE450, OS is Solaris 2.6.

>- Occassionally, on reboot, the non-rootdg group (yantra) disappears. >
>The bootup has messages of the form: "Cannot stat
> /dev/vx/rdsk/yantra/vol01 - ignoring" for all volumes in yantra.
> /dev/vx/(r)dsk do not have any entries for yantra. Sometimes it
>takes multiple (reconfig) reboots for it to be seen again.

Timothy Lindgren says: "This happens when Volume Mgr can't see the A5000's.
Make sure your /etc/system file has an entry such as: forceload drv/socal".

My /etc/system *does* have these lines:

* vxvm_START (do not remove)
forceload: drv/atf
forceload: drv/pln
forceload: drv/ses
forceload: drv/vxdmp
forceload: drv/vxio
forceload: drv/vxspec
forceload: drv/sd
forceload: drv/scsi
forceload: drv/pci
forceload: drv/ssd
forceload: drv/sf
forceload: drv/soc
forceload: drv/socal
rootdev:/pseudo/vxio@0:0
set vxio:vol_rootdev_is_volume=1
* vxvm_END (do not remove)

I'll need to spend some time figuring this out.

Timothy L adds "And NEVER boot or mirror a boot device to an A5000 drive.
That's just asking for problems". I didn't. I think I did not word my
original post clearly enough.

>The root disk is mirrored, but that comes up fine, except for this
>bootup error message: "/etc/sbin/pmconfig: /etc/power.conf: line
>(30) failed to convert mount point /dev/vx/rdsk/rootvol to prom
>name: Invalid argument."

TImothy L - "Since powerd is pointless on a server, remove the startup file:
/etc/rc2.d/S85power".

>
>The A5000 is to be shared with another UE450 in a cold standby mode,
> so when the primary box goes down, we bring up this up with the
> volumes from the RAID (group yantra) mounted.
>
> Since I cannot use vxinstall to encapsulate the disks, how else do
> I create the volumes and the group on this cold-standby? Could
> someone point me to non-contract docs of some kind?

Timothy L -
"You dont need too. All you need to do is "deport" the diskgroup with the
databases to the other hostname, and "Import" the disks on the standby host.
You can do this command-line (preferred) with vxgd -g <dgroupname> deport
(or import). Or the GUI.

Then run "vxdctl enable".

Pre-create all your mount-points, have a stand-by /etc/vfstab file, and
mountall!....

This is a basic senerio in a "clustered" HA environment. Veritas Cluster,
or Sun Cluster automates this process."

I used this info to write a script and ran tests. It worked just great!

---

Mike Martinez pointed me to sunsolve document 18480 that deals with the bootup error message mentioned previously. In a nutshell, this is the power management suspend and resume (CPR) subsystem failing to initialise correctly. Mirroring/encapsulating the root filesystem, as I have done, will produce this failure.

I followed the doc's recommendation and removed the following packages, the logic being that the machine is not a candidate for power management anyway if Veritas/DiskSuite is present:

SUNWpmu Power Management Framework SUNWpmr Power Management Framework SUNWcpr Power Management Suspend/Resume (CPR) SUNWpmowu OpenWindow Power Management SUNWpmowr OpenWindow Power Management SUNWpmowm Online Manual Pages

----

Kruse, Jason K had this to say:

"The only thing that I can think of that is indicated in your message is that the other box has control of the disks when the first goes away. Are you using firstwatch to manage the failover of the disks? If the standby server has control of the disks upon bootup you need to release them prior to the primary server being able to access them. This requires unmounting all volumes, a vxvol stopall, and a vxdg deport yantra command on the standby and a reversal of the procedure on the primary."

I hadn't reached this point at all. My question was more of a "how do I do failover without commercial software?". I just wrote a couple of scripts based on Timothy L's and Jason K's suggestions, which works great!

Thanks to you all for coming to my rescue!

Regards, Rasana

______________________________________________________ Get Your Private, Free Email at http://www.hotmail.com



This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:13:26 CDT