unfortunate sequence of events

Posted by Dick on January 17, 2006

Apologies if anyone’s had trouble with the blog in the last few days.

I recently escaped Stalag MySQL and the postgresql schema load zeroed the id field sequences , so comments (ok, spam) probably weren’t publishable.
Articles certainly weren’t.
I must’ve lucked out on my test post – id 1 must have been free, so the constraint didn’t grumble.

Other than that and a wierd timewarp thing I’m really glad I switched. I’ve always disliked mysql, but on TextDrive there are more good reasons to switch.

  1. you don’t need to supply a password to psql, so you can cron backups in without all the worries about password safety
  2. dedicated db boxes are on the cards. I want to be on the pgsql server with the 2 dozen other guys rather than the mysql server with a bazillion php users (no offence boys, just a load thing).

gruesome twosome

Posted by Dick on October 17, 2005

The more time I spend with them, the more I hate Tweedledum and Tweedledummer

Had another go at updating stitch last night – we just reenabled comments and are getting literally hundreds of spams a day now. Mama used to be the same but it’s been spam-free since getting textdriven last month, hence the urge to move stitch too.

Last time I tried, I stalled on one plugin that refused to play with 1.5. Needless to say this one is essential.

This time I thought I’d clone the site to stitch2, then play with the crash test dummy instance.
I thought this would be simpler than moving it to textdrive and bouncing dyndns records around. Not so.

Mysql and wordpress each had a go at fucking things up.

mysql tries it on: dump live db → test db.

mysqldump --databases livedb | mysql -u whoover testdb -p

mysqldump—databases thoughtfully sticks a ‘use livedb’ at the top of the file, so the above
actually says ‘delete the live db and reload it’. Thank $DEITY for pitiful referential integrity checking or the live site would be a smoking heap.

(you could argue that since —databases is meant to handle multiple dbs it makes sense to switch to the right db before dropping tables. File that one under ‘pilot error’.)

wordpress’ turn: siteurl

So got past that, pointed wp-config.php at the (populated) new db and ran the upgrade. The stylesheet fell off. On closer investigation, all useful links are absolute – clicking on permalinks sends you to the old vhost. If the stylesheet hadn’t moved it would have been hard to spot this. You have to

update stitch_options set \
option_value = 'http://testurl' \
where option_name = 'siteurl';

to make it worth bothering with a test instance at all.

(This one is baffling – php knows what url it was called as, right? If you generate urls based on that, you know the browser can find you again. So leaving aside the unnecassary db lookup, why would you ever want to do anything else?)

nice try, bitches

So at least I’ve got a test site up now. I have resorted to grepping through mysqldumps and the docroot to double check there aren’t any more references to the old domain. I am fully expecting to screw up the old site anyway..

things that make you go ‘bollocks’ (and/or ’sweet’)

Posted by Dick on August 09, 2005

Had one of those weeks where stuff goes pop for no good reason.

First eris, my FreeBSD IMAP/db/Rails/everything-important server screwed its disk (either it’s GEOM in RELENG_5 or the disk itself, but it crashes after an hours use).

Since I’ve started looking at the really very teeny network appliance market lately, I decided now was a good time to get things offsite. Everything is bound to break when I plug in the Linksys anyway, so I might as well do it myself.

First we got assimilated by The Hive. 10 minutes with the exim config and our mail is offloaded there. (incidentally it’s fucking awesome. I’ll talk about that again. )

I got MySQL and Wordpress onto littlebird, the NetBSD wireless gateway, which had a new 1Ghz mini-itx board in it. Found I can reliably kill it by running build.sh tools. This box forwards to gmail, so I need it. Go back to the old board.

But being in the shit is the mother of invention, as they say, so I finally got a $12 Textdrive account to point MX at. Today I finally got round to resurrecting this blog too.

Good stuff so far:

  • The Web interfaces are by and large a joy to use – mysql access and account creation in particular (I’d rather vi my mail aliases as it’s less pain than clicking forms with RSI)
  • you get a lot of bells and whistles for your money. (svn / mysql / lighttpd / full featured mail services / nice development environment )
  • ssh access. lovely.
  • support. I must’ve driven them mad by now, I’ve raised five tickets since the weekend but should calm down now I’ve got the basics covered.
  • security and spam are SEP
  • Less servers on my WLAN. Less noisy fans. Less stuff to break (Anything that escapes ebay == test rigs, and no-one cares if they break. )
  • extras. knowledge base, forums etc. Especially like the RSS status page , that’s a top idea.

niggles:

  • stability. this host has been down a bit, but not for long and I don’t have to fix it.
  • not having root takes some getting used to. It’s odd seeing your mail sitting in the outbound queue but not being able to read it there.