We have been trying to migrate our existing IBM i hosting IBM i partitions to a VIOS hosting IBM i, AIX,Linux configuration. As we have mentioned in previous posts there a re a lot of traps that have snagged us so far and we still have no system that we can even configure.
The biggest recommendation that we took on board was to create a Dual VIOS setup, this means we have a backup VIOS that will take over should the first VIOS partition fail. This is important because the VIOS is holding up all of the other clients and if it fails they all come tumbling down. As soon as we started to investigate this we found that we should always configure the VIOS on a separate drive to the client partitions, my question is how do we configure 2 VIOS installs (each with its own disk) that addresses the main storage to be passed out to the client partitions. We have a Raid controller which we intend to use as the storage protection for the Clients Data but we still struggle with how that can be assigned to 2 instances of the VIOS?? The documentation always seems to be looking at either LVM or MPIO to SCSI attached storage, we have all internal disk (8 SAS drives attached to the raid controller) so the technology we use to configure the drives attached to the raid controller as logical volumes which are in turn mirrored via LVM is stretching the grey matter somewhat? If in fact that is what we have to do? I did initially create a mirrored DASD pair for a single VIOS in the belief that if we had a DASD failure the mirroring would help with recovery, however The manuals clearly state that this is not a suitable option (I did create the pair and install VIOS which seemed to function correctly so not sure why they do not recommend?).
The other recommendation is to attach dual network controllers and assign them to each of the VIOS with one in standby mode which will be automatically switched over should a failure occur on the main adapter. As we only have a single adapter we have now ordered a new one from IBM(we have started the process and it has taken over 1 week so far and the order is still to be placed..) Once that adapter arrives we can then install it and move forward.
Having started down this road and having a system which is non functioning I have to question my choices. IBM has stated that the VIOS will be the preferred choice for the controlling partition for Power8 and onwards, but the information to allow small IBM i customers to implement (without being a VIOS/AIX expert) in in my view very limited or even non existent. If I simply go back to the original configuration of IBM i hosting IBM i, I may have to at some time in the future bite the bullet and take the VIOS route anyhow? Having said that, hopefully more clients would have been down this route and the information from IBM could be more meaningful? I have read many IBM redbooks/Redpapers on PowerVM and even watched a number of presentations on how to set up PowerVM, however most of these (I would say all but that may be a little over zealous) are aimed at implementing AIX and Linux partitions even though the IBM i gets a mention at times. If IBM is serious about getting IBM i people to really take the VIOS partitioning technology on board they will need to build some IBM i specific migration samples that IBM i techies can relate to. If I do in fact keep down this path I intend to show what the configuration steps are and how they relate to an IBM i system so they can be understood by the IBM i community.
We have a backup server that we can use for our business so holding out a few more days to get the hardware installed is not a major issue, we hope that by the time we have the hardware we have some answers on how the storage should be configured to allow the VIOS redundancy and make sure we have the correct technology implemented to protect the client partitions from DASD failure.
If you have any suggestions on how we should configure the storage we are all ears 🙂