Originally posted by: SystemAdmin

Im trying to understand what issues or complications might arise from the following theoretical issue.

Lets say i have 2 x p570, each running dual VIO servers, serving a number of virtualised lpars. The VIO servers are providing both virtual ethernet and scsi to the client lpars. The client lpars are running SAP/Oracle services. Back end storage is DS6000.

In the event of losing the primary p570, would it be feasable to recover my lpars onto the 2nd p570 which also runs VIO from mksysb images via NIM.

Thoughts and comments welcome.
#AIX-Forum

Originally posted by: SystemAdmin

Hi!
I think you dont want to recover your 2 VIOs into the second P570, right?
If not, for sure you can recover your Lpars.
I would already create the lpar into the second p570 and when your first p570 crashes, you just activate and restore the lpar.
But, you must have the same resources (cpu, memory, etherchannel, hba, etc..) to sustain the workload

Regards
Igor
#AIX-Forum

Originally posted by: SystemAdmin

Thats right, the 2nd p570 will have the VIO's configured and running, with standby lpars defined with the appropriate resources.

Only in the event of the primary p570 going down, the partitions on the primary p570 could be recovered using mksysb images stored on a remote NIM server.
#AIX-Forum

Originally posted by: hwyguy

When the primary 570 is running in this example, what is the standby 570 doing?
Does the standby just hold images created on the primary in the event of a failure or is it more of a development environment which can be thrown aside should the primary 570 go down?

Also consider the time involved. If we assume the standby 570 is a true standby and VIOS and LPAR images are refreshed regularly on the standby, then it shouldn't take long to remap your DS6000 LUNS and bring up the LPARS and VIOS on the standby, assuming the necessary cabling (network and SAN) is in place. If not, you'd need to think about the time it would take to recreate the primary hosts on the secondary 570 and weigh that against the amount of time it would take to get the primary back online. And don't forget the time it would take to switch back to the primary 570 if you choose, or you can declare the standby is now the primary.

Another thing you could do is SAN boot the LPARs and VIOS on the primary 570 from the DS6000, and should the primary 570 go down, just re-map the LUNS from the primary 570 HBA's to the secondary 570 HBA's and you'll recover quicker assuming your rootvg isn't corrupt on the SAN.


#AIX-Forum

Originally posted by: SystemAdmin

yes, good food for thought.

Cheers
#AIX-Forum

Originally posted by: steevojb

Hi there,

Considering the investment already made with the infrastructure, a much simpler approach would be to look at HACMP to provide fallover following complete hardware failure.

HTH

Steve

#AIX-Forum

Originally posted by: SystemAdmin

Hi Steve

This solution is being downsized, and the customer is not willing to spend any money on HACMP.

The solution we have found is to unconfigure the luns from the primary VIO, assign them to the standby VIO and map them to pre-defined partitions.
#AIX-Forum