mksysb restore questions?

[polani]

Guys

I do have a scenerio where we have to clone one existing P570 ( POWER5) with dedicated Lpars ( no VIO)to an about to purchase P520 ( POWER 6 ) with virtulaized Lpars.

we have decided to go for mksysb restoration technique , but do have following questions:

1, Will there any be issue while restoring mksysb of a POWER 5 server to POWER 6 server ( AIX 5.3 will be the OS for both source/target servers)

2. Will there any be issue expected while restoring mksysb of a dedicated resources Lpar to a virtualized resources Lpar. offcourse we will be setting VIO with new p520 separately?

Any real experiences to share?

Thanks & Regards



Here comes polani Once again!!!

See my personal blogs at

http://reliablesolutions.blogspot.com

for unique solutions and tips on AIX, Linux,Storage and TSM.

 

[MoreFeo]

1. There should be no problem in restoring from POWER5 to POWER6. I've already done it and I've had no problems. But it would be a good idea to have the base AIX installation CDs, at the same level that your lpar.

2. Only problem here will be the ent devices. Are you going to virtualize thru VIO or IVE? In any case, the ent device will be different, and you'll probably need to reconfigure all your IP settings

 

[markuscj]

I have recently migrated LPAR's from Power5 (dedicated resources) to Power6 (VIOS clients) using mksysb with no problems.

The only difference to your scenario was that the LPAR's were upgraded to AIX 6.1 beforehand - this was because we are using TSM and the IBM TSM support team only support 6.1 in a VIOS environment.

 

[polani]

Guys

Many thanks for responses...

It seems that ppl have already tried it without any problem.

In my case , i will be doing virtualization on target p520 THROUGH ive....

oK , changing IP settings on ent devices will not be any issue , as in any case we have to change Ip addresses manually...By the way , i also do have HACMP running on p570 while there will be no HACMP on p520...

My only concern was storage access.Offcourse on P570 Lpars , we do have dedicated Fiber adapters and Ds4800 disks directly attached to internally booted Lpars. On target P520 , things will change and disks from DS4800 will be assigned as SAN booted disks through virtual SCSI( through VIO).Does this change manageable?

Here comes polani Once again!!!

See my personal blogs at

http://reliablesolutions.blogspot.com

for unique solutions and tips on AIX, Linux,Storage and TSM.

 

[MoreFeo]

If disks are directly accessed with dedicated fc card, you cannot virtualize them. If you do so you will lose the data.

Are these disks rootvg disks, or data VG?

 

[MoreFeo]

What I've understood is:
- you have lpar with internal rootvg hdisks, and external data VG hdisks, with dedicated fc cards.
- you want to go to lpar with virtualized rootvg hdisks and virtualized data VG hdisks.

For rootvg no problem if you use mksysb to restore, but with data VG hdisks you will have a problem.

If you assign these disks to the VIO and then virtualize them to a lpar, you'll lose your data.
Waht you need to do is backup/restore, or migrate from direct connected luns to virtualized luns.

 

[polani]

Hi

Actually at current production site

2 Lpars with dedicated Fc adapters and internal rootvg disks.

Now on proposed DR site ( pwoer 6 , p520)

4 Lpars requirement , so we put one VIO

Now main question was related to mksysb restoration. I can go only for SAN boot process for client lpars as internally i only have up to 4 physical disks...and off course these SAN boot disks will come to AIX client Lpars through VIO.

For data disks , we are planning to replicate on disk level ( ERM)




Here comes polani Once again!!!

See my personal blogs at

http://reliablesolutions.blogspot.com

for unique solutions and tips on AIX, Linux,Storage and TSM.

 

[MoreFeo]

For rootvg you won't have any problem, you just need to ensure the virtualized disks are big enough to hold yor rootvg.

For data disks you cannot use non virtualized luns with VIO server. So you won't be able to do remote mirroring between non virtualized luns and virtualized luns.
To do so you'll need to use NPIV capable FC adapters and switches. With NPIV capable devices, the lpar doesn't see a virtualized disk, instead of it the lpar will see the ds4800 luns as if they weren't virtualized.

 

[polani]

Ok,

Many thanks for clarification.

What if i go for dedicated disks on DR site .. i mean , i will use VIO for processors/memory virtualization and put dedicated HBAs for every Lpar.

offcourse i will need one FC HBA for each lpar , but let's assume i can manage it within same p520, then do you think will there be any issue?

Thanks in advnace for help


Here comes polani Once again!!!

See my personal blogs at

http://reliablesolutions.blogspot.com

for unique solutions and tips on AIX, Linux,Storage and TSM.

 

[MoreFeo]

If you can afford dedicated HBAs for each lpar you won't have any problem with the disks.

In this case, I wouldn't use a VIO. I would do boot from san (rootvg disks directly assigned to the lpars), and virtualize ethernet with IVE ports. In this case you don't need VIO for anything else (CPU and memory virtualization don't need VIO).

 

[p5wizard]

Or use IVE to virtualize network access to all LPARs, no VIO needed at all then. See this or google it yourself.

HTH,

p5wizard

 

[MoreFeo]

If your're new to IVE, take a look at

http://www.redbooks.ibm.com/abstracts/redp4340.html?Open

 

[polani]

Many thanks!!!

Infact using VIO would be required in my case , as we are not going to buy HMC and will be using IVM for partition management...


Dont worry , i have used IVE many time so know it very well

Now , as you technical experts are with me on this solution i will take advantage and i will ask another question here As we will be using ERM to replicate data from production to Dr site ( mainly data disks ) so PVID of disks at DR site will be same as that of production site disks.. Do you ppl think will there be any issue in reolicating data in reverse direction through ERM ? I mean if data changed on DR site and we want to use ERM to replicate data in backward ( or reverse direction)???? as DR site will not have HACMP while production site will have HACMP?


Here comes polani Once again!!!

See my personal blogs at

http://reliablesolutions.blogspot.com

for unique solutions and tips on AIX, Linux,Storage and TSM.

 

[Ashokcse]

1. There wont be any issues in migrating the server from P5 to P6 server. Make sure you are on the supported AIX 5.3 technology level.

2. In the case of virtualized lpar, if you are planning to use NPIV, I think the minimum supported level should be 5.3 TL09. other than that you would not have any issues.

Presenting LUNs through virtual SCSI may not be advisable if you have too many luns. you could use Virtual FC using NPIV setup.

For NPIV, you would need 8GB PCI-express fibre channel adapters.

As far as the reverse replication (or in EMC its called personality swap), it should work fine.