[horde] DB backups

[Kevin Konowalec]

So we had a DB problem last night, which caused imp to silently (to  
the admins) fail.  Nobody knew till many hours later that anything was  
wrong but users weren't able to use the service at all in that time  
period.  Turns out we hit a resource limit and mysql couldn't create  
any new threads so the machine had to be rebooted.  It occurred to me  
that despite having a clustered architecture the database is a single  
point of failure in the system.  So I was wondering with other admins  
running horde in large settings are doing?  Are people running with a  
master and a replicated slave?  Or maybe some other mechanism to  
ensure no service interruption if one goes down?

Also, has anyone ever seen anything like this before?  I was trying to  
automate a nightly dump of my horde mysql database and got the  
following:

# mysql -V
mysql  Ver 14.12 Distrib 5.0.33, for unknown-openbsd4.1 (i386) using  
readline 4.3

# mysqldump -usomeuser -psomepass --verbose horde > horde.sql
-- Connecting to localhost...
-- Retrieving table structure for table horde_datatree...
-- Sending SELECT query...
-- Retrieving rows...
mysqldump: Couldn't execute 'SHOW TRIGGERS LIKE 'horde\_datatree'':  
Can't create/write to file '/tmp/#sql_91_0.MYI' (Errcode: 9) (1)

What does that mean?

Am I doing something wrong?