httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Greg Ames <>
Subject Re: 2.0.23 tarballs up
Date Fri, 10 Aug 2001 17:01:58 GMT
"Gonyou, Austin" wrote:

should I assume that you did a graceful restart first?  and are you
checking the number of processes with ps?  I'll assume both, until you
tell me differently.

> One issue that I can see right now, not sure how big a deal this is, but I
> have a 50 processes running, but before, I had 47..They seem to die after a
> few seconds of inactivity. It used to be at 57, then down to 52, now 50. I
> then took ab and whacked my site real quick.
> ab -n 10 -c2
> ab -n 5 -c2

hmmmm, 15 connections isn't much of a whack, but apparently it doesn't
take many connections to do the job either.

> Now I'm sitting at 29 processes. Lets do another graceful restart.
> I saw my threads go from 30->86->60->33 which is where i'm at now. (margin
> of error is +-1)

how are you checking the # of threads? ps? server-status?

> So, has this looked like a problem? Not to me, but odd none the less.
> Especially when I hit it with AB and my thread count decreases so far.

I'm glad to hear that it's not a problem operationally.  yep, it looks
odd.  It could very well be that you are seeing old generation threads
tied up in mutex land.  Hitting the site with a few ab connections or
real browsers helps clean them up.  I've been toying with the idea of
sending a few dummy connections to the server when we detect this
condition, but there always seem to be more important things to fix.

If you really want to know if you have the "idle threads stuck in mutex
land" situation, you can look at it with server-status.  Old generation
processes usually show up with a single "G" on a line, the threads
blocked on the accept mutex show up as "_" on the same line.  Or you can
do "ps ax -O ppid,wchan | grep httpd".  On my Linux box, threads blocked
in mutexes show up with "rt-sig" or some such in the wchan field.


View raw message