It feels like June has been going on altogether too long though, given the number of things I’m working on, it should have flown by.  June is a month that has defied expectations.  I’m taking Network Performance Testing this quarter, against my better judgement, and so far it seems to be a pretty enjoyable class:  identifying the relationship that exists between business and technical objectives, trade-offs associated with different priorities, etc.  The distance model, which I expected to take to quickly, is unfortunately awkward and, although it is no fault of theirs, my group struggles to maintain reliable and consistent communication.  Such is the way of things.

I built my HTPC on Windows 7 (64bit) using an Asus P5G43T-M, a Core 2 Quad, 4GB of DDR3, a simple SAS/SATA RAID controller, a pair of 500GB SATAII drives, and a Hauppage dual-input media center kit.  Still futzing with the config though – I considered a SageTV build on Debian or another Linux kernel, but scrapped it.

‘Mo, who abandoned me (see what I did there?), also left me with a few parting gifts in the form of wireless routers so my neighborhood now resonates with the strength of my wireless network.  I’m going the DD-WRT route and will engineer a RADIUS server for authentication.  You know.  When there’s time.

I started distilling much of my older hardware into 2 primary groups:  keep and jettison.  I’ll eventually get this packrat mentality under control.

I received some really great LPGA tickets, too, with unlimited access to the corporate/hospitality areas for this weekend so expect zero productivity.

I’ve been working on a new virtualization setup at the office – several ESXi hosts in parallel, vCenter in a VM, vMotion, vStorage, Fault Tolerance, and a VM library accessible for loading shelved VMs.

VoIP, Network redesign, and Solaris licensing is really dragging, though.

I was an early Windows XP adopter having been an early Windows Server 2000 adopter.  Segue:  I received a beta copy of Windows 2000 about a year before the official release and I used that copy until XP was available.  So here we are in the year 2010, nearly a decade after XP became my Windows environment of choice (when I needed a Windows environment) and I encounter a unique Windows issue for the very first time – my user profile somehow became corrupt.

This manifested in a strange way – during the normal boot process the XP login screen comes up and all profiles are available and, upon selecting a profile, the login process begins.  However, shortly after beginning the login process, a strange error pops up indicating that login is not possible because the login profile has likely been damaged.

MSKB article 811151 has a couple suggestions but, from what I gather, this isn’t something you so much recover from as you “fix.”  I tried, unsuccessfully, to perform a Windows restore – the initial suggestion – and later created a second user and followed the KB instructions to move all settings to this new user.  I consider myself fortunate that this was a success.  All I lost was a couple hours.

I don’t have cable TV, so I can’t get a hardware DVR from my cable provider, but I find myself in need ot a DVR – so I’m building one.  I got myself an Asus P5G43T-M Pro, an Intel E5570, an X-master chassis, a couple terabyte SATA drives, and 4 Gigs of DDR-3.  I’m still waiting on the PCI-e tuner cards, though.

The software I need to do some thinking about – although I enjoyed struggling with the XBMC I think my primary contenders are Windows Media Center and SageTV.

Over here there you can find a 2 part introduction to one test-taker’s preparation for the GIAC GCFA exam.  It’s interesting if only because I’ve never know anyone to talk about the GCFA prep.  I was unaware that the exam is open-book and open-notes – so you can build your own cheat-sheet and bring it with you.  I’ve directly linked the posts below:

part 1

part 2

I know Earth week came and went already but it kept me busy:  I virtualized or decommissioned a dozen servers and workstations and replaced 11 CRTs with clean LCDs.  The total wattage amounts to nearly 24,000 watts and included 4 SunBlades, 3 Dell Workstations, 2 Sunfire workstations, a Java workstation, and 2 Gateway servers as well as a number of enormous CRTs.  The side effect of this project was a drop in average temperature in one server room of 6.6 degrees.  I used this as an opportunity to perform some much needed cable management as well.

Over the coming week I have a handful of machines marked for decommission and a half dozen new LCDs to deploy.

I’ve been looking into parallel processing lately – one of our products is fairly java intensive and it would be interesting to see what distributed compiling would do for it.

Easter weekend was  pretty low-key and I spent much of Saturday out in the sun. Made substantial progress through “Under the Volcano,” saw “How to train your drgaon,” and did a small amount of spring cleaning.  Sunday was rough – fought a bug all day and the bug won.  Evidence:  well, it is now almost 2:30 AM and I am probably a few hours from sleep – but I remain optimistic.

Waiting on a RAID1 dual-drive external harddrive enclosure – I don’t really need to have, on hand, several enterprise class servers with multiple terabytes of RAID so I am liquidating the inventory.  I’ll be keeping the sub-notebook computers, an external DVD writer, and some harddrives.  The rest I’m going to get rid of.  I will absolutely sanitize any drives I part with and recommend you do the same.  I recently bought a drive from a user on craigslist and was told it had been wiped but when I attached it and took a peak I found 3 Gigs of Spanish MP3s, some .docx files, and someone’s DC++ settings.  I don’t put much stock in 35 passes of alternating complexity but I do believe in one pass of all zeroes followed by a hash that better be all zeroes and a second pass of all ones followed by a hash of all ones.  I think its easy to be sloppy with data because most folks have no idea what it is or where it is, but if I’m going to be precise with one thing its going to be sanitizing a drive.

I bought a 360, for Mass Effect and Fable II, and given the backwards compatibility I retired the classic Xbox.  It is in the kitchen, sulking, sullen with the knowledge that soon it will be sent away to some unknown fate.

If you recently bought a Sun Sunfire 2270 server and tried to install a RAID card you may also have discovered that not only is the chassis very crowded, but if you attempted to fall back to the onboard RAID, Solaris 10 does not recognize the onboard.  Why would you sell a server with onboard RAID with an operating system which you also developed and that cannot see that onboard RAID?

I can’t think of an appropriate analogy but it is very inconvenient.  It leaves me in the position of needing to find an alternative RAID card which fits and cables which can reach somewhere around 40 inches.  Not in love with it.

Awhile ago, while deploying a new Solaris 10 server with ZFS, I saw a couple unusual options when I was assembling my pools – specifically that there was a “dedup” option available.  I quickly began a hasty love affair with Google and the Oracle/Sun website as I tried to assemble as much information as possible.  I assumed, and this would later be confirmed, that the means of identifying a duplicate would require a hash – calculating a mathematical representation of the target’s uniqueness – and that hash is then stored for later comparison.  Because each object’s hash is compared against every other object’s hash a ZFS configuration with deduplication requires a significant amount of resources.  Deduplication can be performed  on the fly or at some scheduled time when the system is assumed to require fewer resources though the former is probably more ideal if only because the latter requires storage space until such time as the deduplication can occur and that storage can be freed up.  You can enable it really easily, too:

zfs set dedup=on <volume name>

If you think that ZFS should be a little more careful with it’s deduplication you can use a slightly different syntax:

zfs set dedup=verify <volume name>

This modification ensures that blocks are compared to all incoming blocks and the hash is not used – just in case you think it’s possible for 2 blocks to contain different data but possess the same hash.  The mathematical possibility of that occurring is, after all, 1 in (2)^256.

Preface:  I yet live.

One of my critical infrastructure components is a SunFire machine running Solaris 10 – recall that I like Solaris 10 and ZFS – but the Solaris 10 build is so very archaic (circa 2005, I believe) and there is no support for ZFS filesystems.  The primary effect is fairly simple and prevents me from attaching an iSCSI array, a component of a new mass storage array, and configure volumes using ZFS.  So this means I’m upgrading to the most recent version of Solaris.  That isn’t actually true – I am, instead, installing the most recent version of Solaris on a SunFire X4250 (haute technologie) and this will also provide the necessary funtion of upgrading this critical machine as well as integrating a new array.  It will permit us to migrate to ZFS from UFS – no more disk integrity checks – something that would have saved me a few hours early in the year.

In the aftermath of that upgrade, after deploying my new array, I will be bringing in a second array, a Samba device, as a part of my enduser DR strategy.  I’m overhauling the entire backup/disaster recovery strategy, really, and may eventually explain what I had, what was being done, and what I did to make the system comprehensive and operational.

One challenge has been that a finite class C subnet in use has filled up and NAT isn’t much of an option for us.  I am going to solve this by moving to a finite private class B internally which will be managed by a FreeBSD router – I completed the install of 7.2, upgraded with portfetch and portupgrade to 8.0 p2, and compiled a custom kernel using my own DUMMYNET and ifstated post.  I’m going to be redesigning the network as soon as this router is deployed early next week.  This network redesign solves a number of problems not the least of which is a new requirement for a unique subnet allocated to the VoIP solution we’ll be deploying.  I’ll probably post the complete solution, give or take a few details, of this network deployment.

I am also, though not a prestige project, in the process of virtualizing about a dozen machines.  Most of these machines use very little in the way of resources and are just sucking up power.  I have a department that requires direct hardware access and their machines, along with any that require SPARC architecture, will be left alone.  I prefer VMs, though, if only because vMotion and other VMWare tools make VM management easier than machine management.  I’m evaluating the functional impact of VMs on my backup strategy.