Seeing light at the end of the tunnel

I have to admit it ... a lot of my hardware, and the software running on it, has been getting really crufty. Possibly the worst is babylon5, my personal workstation, which has been my main workstation since before I has any separate servers on my home network — and thus had to carry a lot of the server duties as well. For a long time, in addition to being my workstation, babylon5 was also my gateway, mailserver, mailing-list server, primary DNS server, timeserver, backup server, Samba server, and SQL DB server.

In about 2001, gdmusumeci gave me a Sun Ultra30, minbar, that became my NFS server and cvs repository, took over primary Samba duties, and became a secondary nameserver. In 2005, I got hold of a pair of Sun Ultra5s; one, nijo, became my primary nameserver with babylon5 and minbar as secondaries, while the other, yama, became my firewall and router. minbar picked up a pair of 711 disk arrays along the way. babylon5 was still doing everything else. A lot of its software was getting really pretty horrible because, although I upgraded software as I could, I couldn’t really spare it long enough to do a clean reinstall, and the cruft and bit-rot was really starting to accumulate.

Last year, I was given two new machines by darthgeek, one of them a dual-Xeon monster with a rack of twelve hot-swap SATA disks. After 3Ware wasted nine months of my time keeping me dangling for a Solaris 10 driver that they never shipped for a card that they couldn’t be bothered to tell me they’d end-of-lifed eight months earlier, I finally got a pair of disk controllers into it (courtesy of darthgeek again) supported by Solaris 10, which I wanted to use on it for two reasons: ZFS, the Zettabyte File System, and the best NFSd in the industry.

Over the past couple of weeks, I’ve:

• learned how to set up Solaris 10 with a ZFS mirrored boot;
• learned to manage a 3.25TB ZFS RAIDZ2 storage array;
• figured out how the Solaris 10 svc service-management system works, and how to extend it to support my own services;
• learned how to use Solaris 10 zones and set up a new primary nameserver running in one;
• migrated NFS and Samba duties from minbar onto the new machine, babylon4;
• built and installed a complete gcc-4.3.3/binutils/autoconf toolchain, a threaded Perl5.10 and most of the gnu userspace tools onto babylon4;
• installed MySql5.1 on babylon4 and migrated all of the databases from babylon4 to it;
• migrated Postfix, Apache2, MailMan, and Dspam to babylon4;
• finally reinstalled a current version of apcupsd (though I'm still having a few communication issues with my UPS, a HP-badged APC SU3000RM2U that should speak APC's SmartUPS protocol but doesn't seem to do so over the cable HP supplied);
• updated several of my own SQL-based tools that assumed mysqld was running on localhost to be able to talk to a mysqld located anywhere on the network;
• and most recently, finally updated to Bacula 3.0.1, now running on babylon4 and backing up most of the machines on the network as we speak.

That, finally, gets all of the network services off of babylon5. babylon5, finally, is once again only a workstation. And that means that once I transfer all the user files off of it and onto babylon4, I can finally erase a disk on babylon5 and start a completely clean new OS installation on it — most likely Gentoo, based on the scratchpad installation I’ve been working on using the other machine from darthgeek. (That machine, if I can get some quieter fans into it than the shrieking banshees currently in it, will then probably end up becoming my new firewall.)

The light at the end of the tunnel is in sight, and this light at least isn’t an oncoming train. Soon, all the cruft and bit-rot will be gone, I’ll have an up-to-date install on my desktop to go along with the fast new server, and I’ll have learned some useful new knowledge along the way.