AIXpert Blog is about the AIX operating system from IBM running on POWER based machines called Power Systems and software related to it like PowerVM for virtualisation, PowerVC for Deploying VM's and PowerSC for security plus performance monitoring and nmon
I just moved our Power 5 based internal to IBM Wiki Webserver to Power 6 p570 which was recently freed up now that we (Advanced Technical Support, EMEA) are mainly working on the Power 7 machines. The Power 6 machine are still very useful for testing beta copies of new software like AIX and the end of year VIOS functions but some features are Power 7 dependent. We are removing the Power 5 machines because we are moving building and to a much smaller computer room and have to move from 15 racks down to 5 racks. We have been a little lazy in keeping older machine longer than really necessary but it has proved useful on some customer projects. Now we are forced to consolidate - it turns out to be a good discipline we are really thankful we have the Power 6 and 7 machines, their advanced functions for running lots of virtual machines. Also disk subsystems are smaller too.
I used the following process:
Factory reset of the machine using ASMI.
Used Systems Director to update the HMC and system firmware - sweet.
Brand new and up to date VIOS pair so I am are Shared Storage Pools phase 2 ready for later in the year. Each VIOS is on a different CEC of the two CEC Power 6 p570.
The now old fashioned mksysb and NIM install route to bring over AIX, system software and it is also the machine room DNS server.
I thought I might as well update AIX to the latest level (currently AIX 6.1 Technology Level 6 Service Pack 5 or 6100-06-05 to its friends!) - I did not go for AIX 7 at this stage but saved it for later. The prime AIX 7 features of MASSIVE virtual machines, and Versioned Workload Partition to run older AIX versions are not needed at present.
FTPed over the data volume group backup files in a fresh set of filesystems for the website contents - I am are using the AIX mirror pool groups to ensure the mirrors are on different VIOS.
I also upgraded the Apache server to the latest from our favourite AIX open source repository (http://perzl.org) - keep up the good work Michael. This also fixed a warning message we have on AIX start-up about the libsmp library, so that was a double win.
Also upgraded PHP - which the open source and excellent PmWiki uses (http://www.pmwiki.org/).
Once complete, I knocked the old machine off the network (ifconfig en0 down), changed the IP address of the new virtual machine to match the old and rebooted (probably not needed but its nice to be sure and it only takes 40 seconds).
Then the testing! The wiki web-server was fine, updates worked and the Domain Name Server could changed hostnames and IP addresses and the changes propagated.
I had practised the move and upgrade to ensured the wiki would run OK after all the upgrades on a disposable virtual machine the day before. I also performed a reboot test: cold halt using the ASMI menu. This means the HMC nor AIX has any warning of the hard stop. Not perfect but near enough to a power failure. Then used AMSI rebooted of the machine to ensure everything auto restarts: machine + VIOSes + virtual machine + AIX + DNS and webserver. I can do all this testing remotely. The DNS is particular important - if we have a total power failure, it is important to get DNS up and running or we can't remotely access much.
Once live we had two major shocks:
It all worked and we were finished by 10:15 a.m. The Wiki was up and the DNS named had synchronised with the big DNS service in the sky! That throw me a bit - I have to think of something else to do for the rest of the day. I even asked a IBM friend in India to check the website was available within the IBM network but off site ... and it was.
The performance jump was
amazing too but then what do you expect from a Power 5 1.5 GHz to a Power 6 4.7 GHz jump in
CPU speed. The website was something like three times faster when I updated and saved a very large web page and we have reduced the memory quite a bit (13 GB down to 4 GB).
This will be the last time using this mksysb/NIM migration method because in the future we will be using Live Partition Mobility to move to newer Power chips and machines. That will take up just 4 or 5 minutes. In fact, we plan to LPM our website during demonstrations it is that easy, quick and safe.
We (OK actually Gareth) have since set up a new backup system to our USB 500GB disk packs - they were not allowed on Power 5 but now offer us a really simple grand-father, father, son rotation of removable disk packs and off-site storage for a our Power 6 (or 7) machine.
We have a pair of Storwize V7000 disk subsystems on order - I was shocked at the price of this new storage ... it is so low and we opted for the larger disks to get 12 TB usable disk space in just 2U of rack space. I will not tell you the current old SAN disks we have ... as its a little embarrassing :-) We have an entire rack of disks systems and a total disk space of just 3 TB. Seems silly when we can buy a 2 TB disk for a PC for a hundred bucks. OK, to be fair they do have I/O caches, the disks can be shared between many machines and the spindle count means they are a lot faster than a single disk. These disks were sold to a very large customer and then returned to IBM 5 years later, when the customer upgraded. They have been busy in the ATS now for a few more years and giving good service. But they are large in rack space terms and small in GB terms. Is there a market for 36 GB disks, when my USB memory key is 32 GB?
I will update you on our experience of the Storwize V7000 when we get them going, thanks Nigel Griffiths