Our internal server that run lp_publish jobs has been experiencing a highload "seasons" usually 1~2 per week, which doesn't directly correlate to specific cron entries.
I'd suggested "shaping" the likelihood of scripts being run from cron based on current loadavg, as e.g.
load==10 -> P=1.0 down to load==21 -> P=0.0, by prefixing them with:
L=$(awk '{ printf ("%0.0f\n",$2) }' /proc/loadavg);test $((RANDOM % 10 > (20-L))) -eq 0 && ...
, but ~lifeless pointed out that he had observed places for improvement in the scripts being run:
- * * * * * ... launchpad/cronscripts/generate-ppa-htaccess.py
- */5 * * * * ... launchpad/cronscripts/publishing/cron.publish-ppa ...
- */20 * * * * ... launchpad/cronscripts/ppa-generate-keys.py ...
- 3,33 * * * * ... launchpad/scripts/process-death-row.py ...
- 59 05 * * 0 ... launchpad/cronscripts/publishing/cron.daily-ppa
(Not a regression / escalated at this point)