Workload Partition (WPAR) - Answers
nagger 100000MRSJ Comment (1) Visits (21219)
This week I spent 4 hours with a customer covering many advanced WPAR topics and took way a bunch of questions that I had to check the answers and ask the WPAR developers themselves to be sure I had the right answers. If the questions were not clear to my customers and I did know initially know the answers then there may be others with similar issues so I thought I would share the answers with everyone.
1) Workload Partitions - What is the road map?
Well, I am not going to make announcements on a blog for sure. If you are new to WPAR, I can see how some are nervous. Are WPARs a "flash in the pan" or an experiment at version 0.9 or are they a long term investment from IBM and here to stay. It might be surprising that Workload partition have been around for 7 years and after the 2 years of beta testing arrive with the first release of AIX 6.1 in 2007. In fact, WPAR along with Roll based Access Control (RBAC) were the two big feature items of AIX 6 - from my memory there was also the Encrypted JFS2 file systems about that time too, oh an memory keys, and RAS and better performance tuning control. They required fundamental kernel changes which is why it was called AIX6 and not AIX 5.3.<something>. With each new AIX Technology Level (TL) release, the WPAR development team deliver new functions based on their long term roll-out and on customer demand. Here is the chart I use to remind us when the larger features arrived.
So what about the Roadmap? - unfortunately for you I like my job. So I am not going to announce what the developers are working on for the Q4 2012 release. We have seen Versioned WPARs support older AIX versions, namely AIX 5.2 and AIX 5.3 and they are an excellent technology to let us run these older AIX flavours in a supported environment and on modern POWER7 hardware. A quick question: What would be the next Versioned AIX um version? No, it is not AIX 4.3.3! Of course, we have had Global AIX 6 running AIX 6 WPARs from the outset but see question 2) for something that could be more flexible. Then there are features requested most by AIX users with WPAR experience. Also see question 8, which could be improved on! I said nothing, right!
2) Can we run WPAR at one AIX TL level and having the Global AIX at a higher TL level? For example Global AIX and WPAR are running AIX 6 TL5, then we upgrade the Global AIX to AIX 6 TL6 and don't update the WPARs with the syncwpar command. Or we delete and deploy to get to the same result.
Nope this is highly not recommended - if upgrading the Global AIX users should "quickly" syncwpar all the WPARs to match. If they mismatch they will be told to syncwpar before IBM support them.
This is similar to upgrading AIX to a new TL but forgetting to reboot AIX to get it running the new kernel. Running the new code on the old kernel is highly not recommended and only a temporary option, for example, for a few hours before you can take the reboot.
3) Can we Relocate from vSCSI to vFC supported virtual disk?- see red arrow.
It is OK, the only restriction is that the physical device needs to be available on both systems with the same device id. Any configuration that supports that is valid.
Note: Real techies when they move a WPAR to a different copy of AIX refer to Relocation - this is what is on the button in the WPAR Manager software. Marketing guys call this Live Application Mobility or WPAR Mobility
4) Background: An LPAR running WPAR using LUNs for the WPAR files completely fails - perhaps the whole machine is down. So the user want to recover the WPAR on a different LPAR and machine. We can recreate the WPAR as the LUN(s) are available and if we know enough details about the failed WPAR (disks, hostname, IP address, routes etc.) or use the WPAR Manager which recalls the WPAR environment and can Deploy the WPAR on a different machine for us. If the user later starts the FAILED LPAR, what would stop the auto start WPAR messing up the disks that are now in-use in the new WPAR copy running on a different LPAR?
5) If someone on the network broadcasts a packet, does every WPAR get a copy?
6) If we upgrade the Global AIX from AIX 6 to AIX 7 can we upgrade the WPARs too and how?
The WPAR command line tool called migwpar supports this scenario. Upgrade the Global AIX and run then run migwpar for the WPARs.
You should not try to use deploy/synchronize to achieve this (syncwpar is use for smaller updates) - so you can move the WPAR from AIX6 to AIX7 and then run the migwpar command.
7) Two WPARs on the same Global AIX - if they send each other network packets does it have to go the external wire (or virtual switch) or does the AIX realise they are aliases and uses some "loop back" type mechanism ?
Yes AIX notices and does not send the packets out to the physical network. for packets between WPARs residing in the same LPAR. This is not true, if the WPARs have explicit network routing configured to different physical networks.
8) When we Relocate a WPAR can we choose the network that the process memory is transferred across? - We are worried it will saturate the user network.
Currently it is not possible, but this is a possible feature in a future WPAR Manager release. Allowing the user to specify a different IP to be used for the mobility operation.
9) Do we know how popular Workload Partition Manger is from sales records?
Yes we do but IBM don't comment on sales of individual products and never from a blog like this. WPAR Manager sales figures show WPAR Manager is popular and sales are roughly tripling each each year.
The Workload Partition Manager includes two very different items:
Systems Director GUI plug-in Workload Manager
the WPAR Relocation feature which is installed on the Global AIX Virtual Machines.
For a different WPAR user
10) The Exploiting IBM AIX Workload Partitions Redbook table 2.2 suggests that: Exporting a Global AIX hdiskN based on a VIOS vSCSI disk to a WPAR was first supported only at AIX 6.1 TL 6, is that right?
Yes, that is correct. I (Nigel) might add that I had been using such vSCSI disks for a couple of years before it was official announced and supported. I did not known it was "working but not supported". Perhaps, there was a testing time issue or some configurations that I never used that had problems before this TL. vSCSI disks are simple at the client end but there are lots of possible underlying disks to test. The VIOS can back a vSCSI with a logical volume on a local disk, or a whole local SCSI/SATA disk, or a file based large file, or a SAN based LUN ( whole or LV) or now a Shared Storage Pool Logical Unit.
Let me finish with a couple of screen shots of the latest WPAR Manager GUI:
Below is the Global AIX and the matching WPARs that are being hosted view:
Below you get the Actions that you can perform on the WPAR by Right-Clicking it. In the Green square you can see the Relocate option for WPAR Mobility, Live Application Mobility, LAM :-)
Until next time, cheers, Nigel Griffiths