Doing your regular backups is one thing, but doing it right is sometimes a quite different story. Especially when you happen to stumble upon some extreme situations - like e.g. a reasonably small (in terms of actual disk size) Subversion repository that has a rather high number of committed revisions.
There are some servers that "Just Work" (TM) (R). The trouble is, that sometimes you just have to take them down...
Some couple of days ago I came across a rather annoying problem - I had a simple Python program that worked perfectly on my computer, while deployed on the target machine - it failed. What turned to be the problem?
For over a year now I've been struggling with something I thought would never get solved - using catchall adresses with Dovecot's deliver (LDA)...
You may say that the subject is over-exaggerated, but see this piece of "automation"...
