mailgraph doesn't work any more?

[c00l.wave]

Happy New Year to all of you!

I use mailgraph on 2 hosts to monitor postfix. For some reason it suddenly stopped working: Host 1 seems to have failed since yesterday at about 9 am, host 2 issued a huge number of broken threads since midnight (year change), some running very long (infinitely?) and messing resources, some just ending up at the rm-command (probably a race condition). I start mailgraph using crontab:

[steveb]

Why do you run mailgraph from crontab? I am runing mailgrph on serval boxes and none of them has a problem. But I use the /etc/init.d/mailgrph script to update the graphics.

[c00l.wave]

I use metalog, so my logfiles rotate and I need to concat them each time mailgraph gets called. At least I didn't read anything about mailgraph monitoring log rotation or something like that, so there probably would be some missing data. Furthermore I couldn't get it running as a different user using start-stop-daemon (funny I ran the cronjob as root too; I must have been very tired when setting this all up).

Anyway, it worked perfectly for nearly 2 months and sadly it starts to mess now. I found a notice about data loss after year change on mailgraph's web page but no information on how to fix this. Furthermore I can't imagine this to be the problem on host 1 which refuses to log since about 15 hours before year change, neither does this explain why host 2 started to spawn that many zombies. What do I run statistics for if I have to reset them each year? Guess I'm going to write some log analyzer myself or switch to something else...
_________________
nohup nice -n -20 cp /dev/urandom /dev/null &

[steveb]

Mailgraph does work even with logrotating. Do you think mailgraph will always parse the complete log? No way! It only parses the difference since the last run.