zomo technology

RT upgrade

Wednesday, 11. 23. 2005  –  Category: all, sw

escalation:

• want to give someone read-only access to a particular RT queue
• on rt 3.2.2 this works, but user and group information is leaked through the Configuration menus
• rt 3.4.4 has a specific permission knob for this menu
• rt 3.4.4 doesn’t like mod_perl 1.99.x
• mod_perl 2.0.2 contains incompatibilities over Apache:: versus Apache2::
• some modules, such as Apache::DBI, pass RT’s dependencies test but are not compatible with recent mod_perl

I found I had to get more thorough about removing Apache-related modules from my perl library tree and then re-install RT’s dependencies. Of course, I did all this in the rwong order…

PostgreSQL replication

Thursday, 11. 3. 2005  –  Category: all, sw

A project called slony1 has a bolt-on replication toolset for PostgresSQL 7.x – it uses a daemon and a set of stored procedures to replicate table data across clusters.

The list of tables and sequences it replicates has to be hardwired (hurrah for introspection) and it needs a whole nest of perl(1) support tools to generate configs but it does what it says on the tin.

I found these notes hugely helpful in setting things up.

sendmail AAAA woe

Thursday, 11. 3. 2005  –  Category: all, sw

In the normal run of things, the moment I want a host’s involvment with mail to be anything more complicated than just sending it I replace sendmail with exim.

Today’s struggle revealed that even setting a ‘smarthost’ (what everyone else calls ‘relays’) can be too complicated. Not entirely sendmail’s fault though, a Win2k server is serving duff AAAA records. Sendmail likes to make loads of name service lookups, and just one of these duff records will result in it sulking:
Nov  3 07:00:37 dave sm-mta[33084]: j9T7biPo056996: to=admin@zzz.co.uk, delay=4+23:22:52, xdelay=00:00:00, mailer=relay, pri=22441893, relay=budd.xxx.yyy.co.uk., dsn=4.0.0, stat=Deferred: Name server: budd.xxx.yyy.co.uk.: host name lookup failure

Now, there isn’t an AAAA record for this machine; but when sendmail queries for it, it reports SERVFAIL, whereas if I dig(1) for it, i get the correct null reply with NOERROR. Even setting WorkAroundBrokenAAAA doesn’t change this.

This seems to be the same problem as here.

I spent about an hour of this. Gah. Seems I’ll be installing exim every time from now on…

Anyway – debug sendmail name resolution over a queue run with
sendmail -d8.20 -qf