The Great HP Virtual Machine MC/ServiceGuard Cluster Experiment
I place system architects into two categories. Former system administrators, and managers. Out of these two groups, only one designs entire infrastructures, wide-eyed, based on vendor's incredible tales of prowess and ability. And any sysadmin who's ever sat in on a vendor new-product briefing has experienced, with disappointment, the awe of watching your manager's toothy grin spreads across his face as he's spoon-fed anything he wants to hear about performance while the sales guy's eyes turn into dollar signs and start rolling around his sockets like a slot machine. Two star-crossed lovers meeting for the very first time.
And if any of it were real...well, that would be awesome.
The truth, however, usually lies somewhere in between. As a system administrator, I used to be wholly responsible for the cradle-to-grave life-cycle of an entire system; hardware and software both. We would receive the system in boxes, and carefully assemble our own interleaved memory configuration, torque our own CPUs, and rack our own servers. It was our responsibility to troubleshoot our own hardware, order it, and replace it, along with all our operating system duties.
But in all of this technology-awesomeness, we've somehow lost our way.
I'm not allowed to assemble hardware any more, nor replace my own memory or CPUs. Many times the vendor won't support a system in which they haven't had "authorized" personnel do the work. Worse still, there seems to be a trend in corporate America to disallow system architects from actually touching the systems. They're allowed to design, yes, but without any tactile feedback. In fact, the supposition is that these so-called architects will be able to design systems solely based upon the vendor claims. See where I'm heading with this? On the other side of the coin - and in their defense, I no longer have 40 boxes to support. I now have one-hundred and forty!
Yet, there are still those among us who thirst for knowledge.
Next week we have a vendor coming in to school us on the pros-and-cons of clustering virtual machines across discrete hardware versus clustering inside virtual machines on the same hardware :/ Because my group has no hands-on with this technology, I got with our clustering expert who furtively handed me his post-millennium student guide in hopes I could introduce him to HPVMs (formerly Integrity Virtual Machines) as he would introduce me to clusters. And where did I get all my hardware? From the system architect who'd like to have just a little real-world knowledge of such an design prior to the vendor showing up. So I dusted off my trusty rx2600, acquired another Sun 711 "six [disk] pack" (I'd given my other one away to mr_dowg's cousin-in-law after seven years of disuse) and...eventually went through EVERY SINGLE SCA HARDDRIVE IN THE HOUSE because...well dude, I learned that clustering requires disks. Lots, and lots of disks.
It took me 72-hours:
Bask in the fruits of my labor.
I'll publish a HOWTO for instant hpvm-cluster gratification once I polish it up.