Your first reaction might be "Why would you do that, Nigel?" and "Isn't this a AIX blog: Why are we talking about Linux?"

Well AIX is great and all that but when it comes to Open Source it is not Linux. Like it or not most Open Source stuff is developed on Linux and there are readily available binary compiled for Linux and 100's of Open Source packages actually come with the the distribution on their DVD media and arch repositories on the Internet for rapid and straight forward install and updates.

This is an area which AIX must improve and I hear on the grape vine there are projects looking into this area. [This is not an announcement :-) ]

In the mean time, you might think if you have a need for say the very latest in Apache2 webserver, PHP, Apache mods, rrdtool, Perl to support some tools like LPAR2rrd and Ganglia then installing this on Linux takes seconds.  OK AIX comes with Perl now but not the rest and the IBM Apache versions are never current and often not compiled with the right modules.

You might think that I should be recommending the official supported Linux distributions like SUSE SLES and Red Hat RHEL and don't get me wrong there are good and rock solid. These official version come at a cost (althought there might be a time limit trial version available - I have not checked recently)and Fedora 18 and Debian 7 are free. Note too they don't come (as far as I am aware) with the extra IBM supplied packages to allow dynamic LPAR changes like CPU and memory additions or reduction, some of the support and diagnostics commands nor Live Partition Mobility (LPM) support. Being free though allows me to download and use them without getting entangled in purchasing, procurement, management processes.  My idea is I could use these for a few low key, systems management function that are definitely non-production like performance monitoring.  If these work well, then I may have the justification for moving to the supported Linux flavours or even AIX.

I have not used these two free Linux version for some time (apart from some nmon for Linux compiling and testing), so how did it go?

First, Fedora 18

I was told this runs OK on POWER7 so I had the confidence to just download and give it ago. Here are my notes ...

• I downloaded Fedora-18-ppc64-DVD.iso (4.5GB) from the regular Fedora download site - using Firefox on AIX :-)
• I copied this to my Virtual I/O Server's virtual optical library
• Then created a new virtual machine (LPAR) with Shared, uncapped CPU with Entitlement of 0.5 CPU and 2 GB RAM.
• I attached a 16 GB virtual SCSI Shared Storage Pool disks - I use these every time now. As Linux will just see vSCSI they will work fine with zero OS changes. I did set this up with multi-pathing to the two VIOS and Fedora worked that out = cool.
• Created a putty ssh session to the HMC and as my HMC user then ran vtmenu, selected the machine and virtual machine (LPAR). 
  This gives you a terminal emulation session roughly 10 million time better than the regular HMC Console Terminal Window which is, lets face it, garbage!    With Putty + vtmenu we get: Cut'n'Paste working, we can make the window larger or full screen, scrolling works, cursors keys and Tab works, no strange control key incantations to Tab backwards around the screen and no hieroglyphic characters or wrongly placed parts on the screen.
• I booted the VM, selected the DVD device (actually the default on new VM's) and hit RETURN to the prompt "boot:"
• Three seconds later it crashes to the Open Firmware prompt - oh dear, me !!

Enter "help" to get some basic usage information
boot:
Please wait, loading kernel...
Elf64 kernel loaded...
Loading ramdisk...
Claim failed for initrd memory at 02000000 rc=ffffffff
ENTER called ok
0 >

• So I looked around the Internet and IBM DeveloperWorks Linux wiki for hints on Fedora install and found very little to be honest. Then it occurred  to me that Fedora is strongly related to Red Hat - as far as I can workout it is the experimental fore runner for Red Hat technology.  So I searched for Red Hat issues with this error message and quickly found the answer.  The above issue (I think) is with larger kernel images and the initial boot loader offsets need adjusting at the Open Firmware prompt and the boot loader rerun.  I found the information here:
  • Red Hat docs but not very helpful as no example command is shown:
  • https://access.redhat.com/site/documentation/en-US/Red_Hat_Enterprise_Linux/6/html/Installation_Guide/ch-Boot-ppc.html
  • Much more helpful IBM DeveloperWorks blog by Daniel J Martin - Daniel - thinks for the content and clear information.
  • https://www.ibm.com/developerworks/community/blogs/powermeup/entry/rhel_kernel_image_to_large?lang=en
• At the Open Firmware I ran:
  • setenv real-base 1000000
  • reset-all
• reset-all - is one to remember - this launches you back in to SMS again :-).
• This time it boots fine and I continue with the installation - it seems to install the minimum at this stage as the whole install took barely 2 or 3 minutes. It only asks for a Timezone (you select from the list), root user password and the disk layout (I just took the simple one whole disk layout option). Then it installs and finishes with "Hit RETURN to reboot", it set the boot list for me, and we get to the Fedora 18 login: prompt 15 seconds later.
• Now we are in Linux land - it is not on the network after the install - this is the first thing to sort out and then the date/time.  I want to set up VNCserver and do the rest of the setup and software install graphically. Thee are all normal Linux system admin tasks.

Second, Debian 7

With more my above experience behind me I thought to give Debian a test too. Here are my notes ...

• I downloaded debian-7.0.0-powerpc-DVD-1.iso, 3.4 GB, again from the regular Debian Download website. Then added to my virtual optical library as before.
• Created a further Virtual Machine just like that for Fedora 18 above, Shared, uncapped 0.5 CPU and 2 GB of memory.
• Allocated a further 16 GB SSP disk but note: I found multi-pathing confused the Installer as it thinks it has two different disks and the final set-up of the boot loader failed. I tried all three boot loader variations(yaboot, Prep, Grub, etc.). So I removed the connection from one of the Virtual I/O Servers = so it was now single path.
• I also found that the bootlist was not automatically setup but that takes 5 seconds in SMS to correct.
• Debian asks a few more questions during the install but nothing hard or unexpected.
• I used the manual network set-ups options (no DHCP for a server) and to get updates from the Internet from one of the standard mirrored archives via a proxy server. Gareth only set this up the same morning - good timing mate.  Doing this network option, right from the outset means updates are pulled down immediately so the install is right up to date and apt-get is also ready to install anything not on the DVD.
• When asked, I did not check other CDs as it was the DVD. I have not downloaded a duff .iso image for years - perhaps one of the benefits for working for a large company that was on the Internet so early on. All my machines are on the whole IBM owned 9.x.x.x networ
• I again used the automatic disk partitioning settings. It all installs without any further issues.
• As it is on the network already that will save time and just like the previous Linux it is all regular Linux systems admin work from here on.

So apart from a half hour looking for the setenv trick these were just "load and go" installations. I hope you find this useful and might encourage you to give it ago and run some experimental Open Source tools or projects.  Cheers, Nigel Griffiths.