Generally speaking, we host our mercurial repositories using SSH; sometimes of course, we need to do it via HTTP because we don’t want to give external contractors SSH access because who knows what damage they can do with a terminal (anyone who’s seen someone do rm -rf * in /etc knows what I’m talking about). It’s bad enough they damage the contents of the repository with their inability to read a good primer site like http://hginit.com (you know who you are).

A quick update this week; I updated the kernel on one of my CentOS 5.6 images and rebooted; it wouldn’t boot as it couldn’t find any LVM volumes. At least you can always reboot into a previous kernel.

If I’m being honest, just blithely running yum -y update, reboot was a trifle silly. Still, if you can’t do that with virtual machines when can you do it (that’s what snapshots are for). The affected machine itself doesn’t really do much other than run a couple of JMS Brokers/MySQL/Postfix/IMAP/FTP/SSH so my unit tests have things to work against (my life is more interesting that this, honest).

I use Cygwin all the time, if it wasn’t for Microsoft Outlook I would have probably given up the Windows platform a long time ago. That there isn’t a PIM out there that is as good as Outlook is a really damning statement in some ways, Outlook isn’t that good if you don’t have an Exchange environment. No, Google mail + calendar is NOT THAT GOOD.

I’ve always been a fan of virtualising my development environment; nothing quite like carrying around a 220Gb disk image of Windows 2003 + SAP R/3 and writing a new SAP connector when you’re on the road. Recently though, I’ve been getting less than stellar performance from vmplayer / vmserver; so I wanted to switch to a Type 1 hypervisor…

As it happens, due to various reasons (I say various, but I mean one reason) I have inherited a couple of laptops, all with broken screens; other than the broken screens they’re perfectly servicable, and yet not worth refurbishing. They’re both Core 2 Duo (T7250 or better)with 4 GB RAM and 250Gb hard drives; which should be more than adequate to run a Type-1 hypervisor; they have Intel-VT and all that. I’m trying to find out if I could install ESXi onto these laptops; shall we just say that there’s a dearth of information on the VMWare site about compatibility with laptop models, so I’m left with one choice, Hyper-V from Microsoft. Why didn’t I just try and install ESXi? I could have; we’re in the process of virtualising all our servers using VMWare already, so obviously there’s a wealth of knowledge inhouse I can tap up, but that wouldn’t have been interesting.

What’s the most important thing when performance tuning the adapter; it’s information, having a gut feeling about where the adapter is slow is all well and good, but you’ll need to prove it. After all premature optimization is the root of all evil.

In addition to having Perf4J (http://perf4j.codehaus.org) annotations on Workflows, and Producers (which can be enabled using aspectj AOP); 2.7.1 introduced the a new service Perf4jTimingService that can wrap any arbitrary service and gather performance metrics about that service’s throughput.