"It runs on what??"
WTF
This rather sub-par domain began in...uh...2001 I believe. Why? Don't really
know why. I guess it seemed like a good idea at the time. It's all the fault
of this guy Raistrick, indirectly, and Andy, more directly (because he burned
the Debian CDs for me). Long story short: I started hosting my own email
server using dynamic DNS (dyndhs I think) on ... dialup. Yes. And later on
ADSL. By that point I was dependent on it, and it was getting annoying keeping
DNS straight and staying online. So this crap was born.
The name of the domain was chosen because, indeed, the servers were all
pretty much built out of glorified piles of garbage (see below). We couldn't
get anything to run reliably because of assorted hardware problems. This
eventually improved, but of course the name can't really be changed so...there
you go, failsure.net.
There are actually other domains hosted here now, as well, but the anchor is still failsure.net. "Anchor" is really an apt term for it, too.
History
- nanbara, K6-2/400MHz. 64MB memory. Clock ran way fast so synced up
using NTP every five minutes to keep everything from going haywire. Floppy
controller busted and couldn't be disabled, so a custom Linux kernel without
any support for floppy controllers was needed or the machine would crash.
Flakey CD-ROM drive made initial installation a challenge. Motherboard was physically damaged (cracked). But...it worked enough for us. We actually had a
text file with a list of broken stuff on the thing to remind us. Once ran
badblocks on one of the drives for about two weeks before giving up on it.
Hosted remotely (me in GA, box in CA) at Harvey Mudd.
- oji, PII 400. 128MB memory. This machine was a lot better. No real
hardware issues. Still at Mudd.
- dilandau, dual Pentium 66. 160MB memory. This thing was a beast. Would
heat an entire room, no lie. Had 7 ancient SCSI drives in a RAID-0 on a single
controller (hey, I didn't know any better at the time). Miraculously never
suffered a drive failure or data loss. Had an EISA bus. Hosted in GA at
home, on Mindspring (later Earthlink) ADSL.
- haruko, Celeron 533. 384MB memory (eventually). This was when I splurged
and bought a 1U case on eBay and threw together a "real" server. That's right -- a "real" server with a Celeron! This was the first time I had actually
spent money on equipment for failsure.net. I could hardly believe it myself.
This thing has lasted a while, hosted: Biltmore Communications, interesting
wireless broadband in the highrise condo I was living in at the time. Next,
DirecTV DSL at an apartment in Alpharetta, GA. DirecTV decided to get out of
the DSL business, and I didn't want to pay $15 a month extra to Earthlink for
a static IP so...crisis! Averted by moving hosting to my new place of
employment. Thanks Randy!
- UPGRADES: Upgraded to 1GHz PIII. Worked for a while and then the caps
burst on the mobo :-( Nice while it lasted...CPU was free out of a dead
machine.
- haruko, Pentium II 350...temporary. 256MB. Using this as a temporary home
for failsure.net until I can replace the mobo in the 1U chassis. I don't
want to spend any money...
- haruko, Celeron 533...again! Discarded eMachines motherboard with a fat 512MB RAM and the same damn processor and lousy old 9.1GB drives in a 1U case. I even have a spare identical motherboard, ensuring years of below average performance! :-) 233 days uptime as of this writing, not bad for a pile of junk.
- haruko, Pentium III 933, Compaq DL320. This was in a big pile of returned equipment from a customer at my previous employer. I swapped a 386 I had been hanging onto for it (this was totally ok -- hard to explain here, but part of a vendor buyback program...they would take anything). This machine had been through hell. It's all scratched up and beat up, and the casing on the right front is broken off. But it's the best box yet - 1GB memory, 2 18GB SCSI drives, real server hardware!
- haruko/kos-mos. haruko is now virtualized on another server to save power at the house. Using Linux Vserver to run it. Server is a recently upgraded dual Opteron 252 (2.6GHz) with 4GB memory and, uh, 2.7TB of storage.
- haruko, nanbara's old hardware. Decided to end the virtualization of failsure.net by hosting it on old junk. K6-2 400 @ 500MHz, undervolted, 3 HDs with no RAID just a bunch of logical volumes, 256MB RAM. Oh yeah.
- haruko, back on the DL320...it has 2GB now.
- Back on nanbara's hardware. This time, have about 15GB usable mirrored space, comparable to the DL320. Except for speed. But it's usable.
- Switched to naga, Sun Blade 100 with NetBSD. Just to mix things up.
- Back on nanbara again. NetBSD was fun (six months) but I loves me some Debian.
- "new" hardware, synchro: 1.4GHz Celeron wedged into a Slot 1 system with a funky PowerLeap adapter I've had for ages. 768MB ECC memory and 2x40GB drives. Gave away all the ancient hardware I had -- a PIII-based Celeron is still pretty old...
- Decided synchro would work better as a file server, so took the Sun Blade 100 I was using for that purpose and built it into a web server. The machine was renamed to t-elos some time ago. 500MHz USII, 2GB memory.
- Moved to a VirtualBox VM on synchro, allocated 128MB memory and 10GB disk.
- synchro retired, moved the VirtualBox VM to kos-mos, upped RAM allocation to 256MB.
Failed components
Before I started keeping track: various hard drives and cooling fans. Listed in order.
- Broken pins, PowerLeap CPU upgrades on dilandau. Not really the hardware's fault. User error.
- 9.1GB IBM 50-pin SCSI HDD in original incarnation of haruko (one of three).
- Burst caps, Socket 7 mobo on haruko...caused replacement w/found eMachines crap.
- ISA NE2000 NIC, couldn't find jumper settings for it. It's ok I found another one!
- ISA 3c515 NIC, kept resetting with heavy HD activity...
- Quantum "Bigfoot" 8.0GB IDE HDD, the big ol' 5.25" jobs. Yep, it died.
- Quantum "Fireball" 2.5GB IDE HDD. Listen, they don't make Quantum HDDs anymore, and I'm starting to see a trend. On the upside this thing made really cool noises when it failed.
- Cooling fan seized on the DL320. Was on the power supply; when it stopped, the machine stopped. Fixed with Liquid Wrench(tm). Still works!
- While rebuilding nanbara (socket 7 machine) a 15.3GB Seagate drive failed. This one was really interesting. You could get it to work for a while by removing the molex connector after the machine was powered on, and reconnecting it. I actually did try a couple times to get it to work, but eventually gave up.
- While building synchro found a bad PC100 256MB ECC DIMM using memtest. There were two "stuck" bits, always bad in the exact same spot.
- While building synchro, 80GB Maxtor/Diamond (reman) IDE HDD that made cool click/thunking sounds but otherwise failed to function.
- ...and a day later, a 40GB Western Digital (reman again) IDE HDD failed while in the process of getting t-elos setup. RAID. EVERYTHING.
- 250GB SATA Seagate drive. Lasted 7 years, can't really complain.
Yet to fail
Never had to replace a power supply (cooling fans IN power supplies, yes) or CPU. Not that these haven't failed in other systems!
server statistics
Last updated 2013-02-26