February 7th, 2009



Boy was suffering from a variety of infections due to a decreased immune system brought on by the flu, which he apparently contracted without us being aware of it. The doctor put him on bedrest for the entire week and I worked from home - for a week. A very long week. I've been hopped up on MONSTER and scooby-snack, and last night I was dancing the Numa Numa Dance along with that chunky kid from NJ for hours on end after bsdcat sent me the URL.

Usually, I don't like working from home unless I have something to do, so I begged for work, and it was given to me tenfold. Perhaps more than I wanted to do, nonetheless, I managed to accomplish everything by Friday afternoon. I had to get a environment ready for production. Fortunately the machines had already been built, so all I was doing was readying them. It was pretty basic sysadmin stuff, but the proof, as they say, is always in the pudding. No, the manually intensive part boiled down to access:

  • One terminal concentrator

    • Seven ports

      • Two domains per port

        • Half a frame on each port

          • Six servers on each frame

Each requiring a root passwd reset (I was provided a list of 50 possible previous root passwords to work from) and thus began my access hurdle. I knew I wanted tabbed terminal windows for this activity (opening each frame's six servers in a single window) and PuTTy is a touch behind the times on this, and my work laptop which is a dual-monitor setup running ubuntu doesn't do well on my awkwardly aspected 16:10 widescreen home monitor. So...I turned to OSX which handled the job flawlessly. That is, once I found a VPN client for OSX, imported the settings, and killed all the CPU-sapping processes (telnet [to my HP/UX box's (hereafter referred to as belanna) Management Processor] had apparently hung driving utilization to 100%). But then that's why I have a KVM, and my stalwart XP box - for the heavy loading. I really would like one of those new Psystar boxes. Oh the wonderful things ehowton could do!

New 320MB/sec SCSI drives in the previously empty drive slots

Ever since harddrives went over 9GB, I stopped breaking out discrete filesystems, a practice which has served me well in Solaris. Not so much in HP/UX. My 11iv2 (11.23) installation filled up, just prior to that feudal hardware-excreting lord Ernest asking me to perform an 11iv2 to 11iv3 (11.31) upgrade out of curiosity to see if the full-version VxVM which was included on the "Mission Critical Operating Environment" (MCOE) would be retained. Because this coincided with work's decision to start rolling out and testing 11iv3, rather than repair the existing filesystem problem prior to performing the upgrade, I just installed 11iv3 standalone, and was surprised to discover in the "Data Center Operating Environment" (DCOE) a plethora of installations options ranging from a full-scale High-Availability Suite, to a workstation-level install ("Technical Computing Operating Environment" (TCOE)) - which not only includes the full VxVM, but now also bundles their Auto Port Aggregate (APA) NIC-load balancing/failover software. There are probably many other packages included that I'm not yet aware of which were previously unbundled, but for me, the sweet spot was the inclusion of the VSEOE, or "Virtualization Server Operating Environment" which not only fully supports virtual partitions (vPars) and virtual machines, but 11iv3 now takes advantage of a previously unknown feature of the Itanium 2 (ia64) architecture: Hyerpthreading! In theory (and I'm a long way from practical application) I would be able to carve out a linux instance, running within HP/UX, and assign it a portion of a processor.

Here's where things get dicey. In Solaris, any machine which can run Solaris 10 is capable becoming a host for other Solaris "zones" or virtual servers. With HP/UX, only specific hardware is supported. Now I'm no fool and learned a long time ago that 'supported' is merely a term used by companies when they no longer want to technically sustain an application. In other words, just because its not supported doesn't mean it won't work. HP/UX vPars however, while different from nPars (electrically-separated servers in the same machine) are not "zones" as in Solaris. They require a specific 'frame' in which to reside, and apparently the rx2600 *actually* won't run them. But I'm still running that down. Also, the VSEOE's "virtual machine" being mentioned outside of vPar leads me to believe that a Solaris zone-like build is now possible.

Lastly, SAM has been almost entirely deprecated, having been replaced with SMH - System Management Homepage. As I'm running this headless, I've only seen its cli interface. If you can believe it, its actually slower than SAM, but does seem to have a more intuitive interface.

One last story about Ernest. Today I rec'd an invitation to "connect" with this attractive lady on Yahoo:

Yet she has no profile and I don't know who she is. How did she find me? What does she want with me? I'm all for expanding my friend base and interacting with new people - I actively seek it out - but with no note or clue as to who she is, or why, frankly I'm baffled.

Which brings me to Ernest. Awhile back his wife found my Facebook page and invited me to join her network. Before I could accept, however, I was directed to chose from a small set of circumstances which explained how I knew her. Unable to proceed without this data, and without an accurate choice for depicting our relationship, I chose what I thought at the time was the most ambiguous explanation, "We hooked up."

Apparently in the parlance of our times, that was the wrong box to check.