Comment 10 for bug 802850

Tried with smaller amount of workers and removed the setTimeout() call from worker startup code. Restarted all workers and started 3 cpu-intensive tasks. Only 1 gearman server, no persistence, running as root.

2 workers handled those 3 tasks, resulting in a 20-30 avg load in a few minutes (encoding video). As subprocesses were "niced", machine handled that load quite nicely. After about 2h of continuous work, server stopped responding to pings, http and ssh ports timed out, load skyrocketed above 40+. Other workers started to spin at 100% cpu and the whole thing became unresponsive for about 20 minutes :-(