Exporting VVM diskgroups before jumpstarting to Solaris 10
I have a Solaris server that is currently running Solaris 8. The server has attached SAN storage with data on them. I want to export the diskgroups that are on the SAN as to (hopefully) protect that data. Then, I want to jumpstart the server placing a new image of Solaris 10 onto the internal drives. Finally, I want to be able to import and mount the SAN storage drives that were exported before the jumpstart. I should also mention, the current config that is on Solaris 8 is utilizing Veritas Volume Manger 3.5....our jumpstart server places Veritas Volume Manager 4.1 along with permanent licenses on the server....so I will be exporting the diskgroups under VVM 3.5 and Solaris 8 and then importing the diskgroups under VVM 4.1 and Solaris 10. Any command line suggestions for doing this properly both with the export and then the import would be greatly appreciated. I should mention that I don't work that frequently with VVM and was thinking of using vxdiskadm for doing the export and import, but any suggestions would certainly help.
# 1
I'm not sure you really need to do this - as any active imported volume groups will be stopped and deported when you shut the machine down...
if you're looking to take backups of things - do the standard tasks from something like explorer's /opt/SUNWexplo/tools/vxvm script
probably don't need to run all of these, but if you're trying to be safe -
/etc/vx/diag.d/vxprivutil dumpconfig ${PRIV_PATH} \
/sbin/vxlicrep
/sbin/vxlicrep -e
/usr/bin/ls -lLR /dev/vx
/usr/bin/ls -lR /dev/vx
/usr/sbin/vxcmdlog -l
/usr/sbin/vxddladm listexclude all
/usr/sbin/vxddladm listjbod
/usr/sbin/vxddladm listsupport all
/usr/sbin/vxddladm listversion all
/usr/sbin/vxdg -g $DG_NAME free \
/usr/sbin/vxdg -q list
/usr/sbin/vxdg list $DG_NAME \
/usr/sbin/vxdisk -o alldgs list
/usr/sbin/vxdisk list
/usr/sbin/vxdisk list ${DISK_NAME} \
/usr/sbin/vxdisk path
/usr/sbin/vxdmpadm getdmpnode enclosure=$enclosure \
/usr/sbin/vxdmpadm listctlr all
/usr/sbin/vxdmpadm stat restored
/usr/sbin/vxprint -Ath
/usr/sbin/vxprint -h
/usr/sbin/vxprint -hmQqg $DG_NAME $VOL_LIST \
/usr/sbin/vxprint -hmQqg $DG_NAME \
/usr/sbin/vxprint -hr
/usr/sbin/vxprint -mdg $DG_NAME \
/usr/sbin/vxprint -mvrGg $DG_NAME \
/usr/sbin/vxprint -rmvg $DG_NAME $VOL_LIST \
/usr/sbin/vxprint -th
/usr/sbin/vxprint -thrL
/usr/sbin/vxprint -vng $DG_NAME \
/usr/sbin/vxtranslog -l
# 2
Ok, prior to the jumpstart, my SAN disk output for vxdisk list looked like this:
c2t0d60s2slicedq3ivr1m1_dg-01 q3ivr1m1_dg online
c2t0d61s2slicedq3ivr1d1_dg-01 q3ivr1d1_dg online
I jumped the system and was able to import the diskgroups above, however, they now look like this in vxdisk list:
emcpower0s2 auto:slicedq3ivr1m1_dg-01 q3ivr1m1_dg online
emcpower1s2 auto:slicedq3ivr1d1_dg-01 q3ivr1d1_dg online
If I try to remount them (after making the directories), I get the following error:
UX:vxfs mount: ERROR: V-3-20003: Cannot open /dev/vx/dsk/q3ivr1m1_dg/q3ivr1m1_vol1: No such device or address
UX:vxfs mount: ERROR: V-3-24996: Unable to get disk layout version
So, first, I need to find out why the name of the disks changed and secondly how I can get them remounted to the server.
# 3
have you checked the veritas support website to see if there are any ASL packages you need for your specific storage platform support?
I'd suggest going to support.veritas.com and digging around there - if you know the vendor (assuming EMC) and model of the storage frame you're attached to...
at this point, I don't yet believe it's an issue with disk names which is preventing you from using the storaage.
btw - you mention name, my guess based on what you had listed is that you may have had EMC storage and were using native names, or using dmp, or using mpxio prior to upgrade, and now you're using powerpath
seeing that the disks are showing in a vxdisk list output means you have them visible in format, so my current guess is that you need the ASL
# 4
Ok, I am a little further ahead now and wanted to post what I did in case someone else encounters this. To get the disks imported again, I performed a vxdg import <diskgroup>. I could not mount the volumes until I discovered I needed to run the command: vxrecover -g <diskgroup> -sb . After running that command the volumes were started and I once again was able to mount them up. Now as for the change in disk names, I will take your suggestion and scout around on the Veritas site to see if I can find any info there. Thanks for your suggestions and your help.
# 5
> Ok, prior to the jumpstart, my SAN disk output for
> vxdisk list looked like this:
>
> c2t0d60s2slicedq3ivr1m1_dg-01 q3ivr1m1_dg
> online
> 2t0d61s2slicedq3ivr1d1_dg-01 q3ivr1d1_dg
> online
>
> I jumped the system and was able to import the
> diskgroups above, however, they now look like this in
> vxdisk list:
>
> emcpower0s2 auto:slicedq3ivr1m1_dg-01
> q3ivr1m1_dg online
> mcpower1s2 auto:slicedq3ivr1d1_dg-01
> q3ivr1d1_dg online
>
Yes, the disk names may change with the upgrade. That shouldn't affect the actual structures (like volumes).
> If I try to remount them (after making the
> directories), I get the following error:
> UX:vxfs mount: ERROR: V-3-20003: Cannot open
> /dev/vx/dsk/q3ivr1m1_dg/q3ivr1m1_vol1: No such device
> or address
> UX:vxfs mount: ERROR: V-3-24996: Unable to get disk
> layout version
Since you explicitly imported the dg rather than an autoimport at boot, the volumes are not started. Did you start the volumes?
vxvol -g q3ivr1m1_dg startall
> So, first, I need to find out why the name of the
> disks changed and secondly how I can get them
> remounted to the server.
The name change should have nothing to do with the mount issue. You need to check the status of the volume.
vxprint -ht -g q3ivr1m1_dg q3ivr1m1_vol1
Because of the "no such device", I'm assuming it's not started.
--
Darren
# 6
Hi,
I performed similar tasks of upgrading Solaris 8 with Veritas 3.5 to Solaris 10. Since Solaris 10 doesnot support VxVM less than 4.1, I had to upgrade the existing VXVM 3.5 on Solaris 8 to VXVM 4.1. All my diskgroups and volumes were perfectly okay. Then, I went on to upgrade Solaris 8 [with VXVM 4.1 now] to Solaris 10. After upgrade, my Solaris 10 box got stuck up at the point when VXVM initializes the DISK array. I faced the similar problems as you faced.
I seeked help of SUN, and they advised me to uninstall 4.1 and re-install it on Solaris 10. Then I did a "vxdg -Cf import DISKGROUP" and the vxvol -g DISKGROUP startall worked fine with the same volumes which I could mount using the earlier /etc/vfstab file .
May be an uninstall and then an install could solve your problem. Correct me if I am wrong..
Regards
Manoj Achan
# 7
> I performed similar tasks of upgrading Solaris 8 with
> Veritas 3.5 to Solaris 10. Since Solaris 10 doesnot
> support VxVM less than 4.1, I had to upgrade the
> existing VXVM 3.5 on Solaris 8 to VXVM 4.1. All my
> diskgroups and volumes were perfectly okay.
If your root disk is encapsulated then you have to do that. If it's not, then you can do it all at once.
Deport DGs
Disable VxVM
Upgrade Solaris
Install new VxVM
Import DGs
--
Darren
