RE: [sid-milter-discuss] RE: CPU usage

[<fre...@va...>]

>Earlier you posted a truss of the filter that showed it in a pollsys()
>loop.=20
>Can you add timestamps to that truss?  I can't tell if it's spinning or
if >those=20
>are just periodic polls.
>
>However I may have found the cause of the crashes.  They have to do
with=20
>handling of timeouts when multiple queries are pending.  I'll have a
patch >for=20
>you to try shortly.  The bug is specific to sid-milter since it's
caused by >the=20
>way sid-filter is using libar.  dk-milter not be affected.
>
>I haven't managed to reproduce any CPU usage issues with any of the
>filters, but=20
>it could be related to this (and, in fact, fixed by the same patch).

It's only sid-milter that dumps core, so that what I'm worried about.
sid-milter can be "hogging" the CPU for a long time...which in the end,
results in a core, and a restart of the milter...

That dk-milter spins CPU some time is not equally important. It timeouts
in some way, and goes back to normal load. (CPU hogging isn't as bad as
sid-milter).

Okay, just send me the patch and I'll try it out...

/P