Turns out the BigBrother port tests bbnet does were causing this
behavior.
 
Tomcat logs these when BB telnets and quits on port 7009:
 
SEVERE: BAD packet signature 29045
71 75 69 74 00 00 00 00 00 00 00 00 00 00 00 00 | quit............
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 | ................
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 | ................
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 | ................
 
I'm not sure why this caused the curl calls to accumulate Apache
threads, but eliminating the BB check solved the issue.
I'm looking at better ways to make sure tomcat is up.
 
Thanks,
 
Fletcher Cocquyt
Senior Systems Administrator
fcocquyt_at_stanford.edu
 
-----Original Message-----
From: curl-users-bounces_at_cool.haxx.se
[mailto:curl-users-bounces_at_cool.haxx.se] On Behalf Of Fletcher Cocquyt
Sent: Tuesday, June 29, 2004 9:52 PM
To: curl-users_at_cool.haxx.se
Subject: Curl leaving lots of Apache threads around - why?
 
We employ curl (curl 7.10.4 (sparc-sun-solaris2.9) libcurl/7.10.4
OpenSSL/0.9.7a zlib/1.1.4
) heavily to monitor our different webapps (apache 2.0.48 + tomcat
4.1.30)
 
The command looks like:
time curl -s -netrc-optional -k -m 25 -i {URL to monitor}
 
I run the script every 5 minutes and graph the latency in cacti.
 
The problem is, the apache threads associated with these curl calls do
not disappear - but stick around in "sending reply" mode - as reported
by the apache 2.0.48 extended status stats.
 
I know the URLs are actually returning and not timing out because the
graphs would reveal timeouts.
I am puzzled why curl is exhibiting this behaviour with apache.
Eventually I have to HUP the apache server to clear the threads.
 
Thanks for any insight!
 
Fletcher Cocquyt
Senior Systems Administrator
fcocquyt_at_stanford.edu
 
Received on 2004-07-01