httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Cliff Woolley <cliffwool...@yahoo.com>
Subject Re: tarballs are up
Date Tue, 31 Jul 2001 04:07:25 GMT
On Mon, 30 Jul 2001, Greg Ames wrote:

> Could you do me a huge favor?  attach to the parent with gdb, set a
> breakpoint in perform_idle_server_maintenance, and see what you have for
> idle_thread_count at this line:
>
>    if (idle_thread_count > max_spare_threads) {

I haven't done this yet, but that's my next step.  Here's some new
information, though:  I've definitely isolated the problem.  It happens
whenever I do two gracefuls in such rapid succession that one hasn't
finished before the next one starts.  That kills it every time.  I haven't
tried, but I bet if I let it churn for long enough it would resurrect
itself.  The scoreboard gets into a sequence that looks like this (each of
these blocks is a separate call to /server-status):

................................................................
_G......................._________________________..............
................................................................
.........................................G......................
_______________________W__________________________..............
G......................___________________________..............
................................................................
................................................................

................................................................
_G.............................___________________..............
................................................................
.........................................G......................
____________________________CW____________________..............
G............................_____________________..............
................................................................
................................................................

................................................................
_G................................._______________..............
................................................................
.........................................G......................
________________________________CW________________..............
G................................_________________..............
................................................................
................................................................

................................................................
_G.....................................___________..............
................................................................
.........................................G......................
____________________________________CW____________..............
G...................................._____________..............
................................................................
................................................................


I didn't try it, but I'm willing to bet that If I let all of the extra
workers die off in the other two processes (and maybe gave up my keepalive
in that middle process), that the server would come back to life.

More details to follow...

--Cliff


--------------------------------------------------------------
   Cliff Woolley
   cliffwoolley@yahoo.com
   Charlottesville, VA



Mime
View raw message