Thanks for the fast answer.
How many workers would you suggest? I tried it once with 6 and 10 but
because of the memory consumption (swapping) the server became even less
responsive.
Right now we are using 4 workers.
Couldn't it be that this aggressive waiting strategy could even prevent
the child processes from doing their work by stealing them cpu time?
What could it be that keeps a worker busy for more than a minute in the
same request? Is it possible to find out in what requests are taking
most of the time?
Thanks
Raffael
On 07/18/2011 02:34 PM, Ludovic Marcotte wrote:
On 18/07/11 07:13, Raffael Bachmann wrote:
From time to time (sometimes twice a day) the parent process consumes
full cpu.
That happens when all sogod workers are doing some work - the parent
process will spin very quickly trying to find an available worker.
Increase the number of sogod workers.
--
[email protected]
https://inverse.ca/sogo/lists
