general/8725: Processes in keepalive state hang around indefinitely

[Melissa Evans <tr...@mebtel.net>]

>Number:         8725
>Category:       general
>Synopsis:       Processes in keepalive state hang around indefinitely
>Confidential:   no
>Severity:       critical
>Priority:       medium
>Responsible:    apache
>State:          open
>Quarter:        
>Keywords:       
>Date-Required:
>Class:          sw-bug
>Submitter-Id:   apache
>Arrival-Date:   Fri Nov 09 09:30:00 PST 2001
>Closed-Date:
>Last-Modified:
>Originator:     trai@mebtel.net
>Release:        1.3.22
>Organization:
apache
>Environment:
SunOS webmail 5.8 Generic_108528-03 sun4u sparc SUNW,UltraSPARC-IIi-cEngine
Using gcc.
>Description:
After a short while of running the server, it is evident that processes are hanging around indefinitely.  Kill -ALRM and Kill -USR2 do NOT work to kill these processes. 

server-status often looks like:
KKK_KK_KKKKKKKKKK_KKKKKKKKKKKKKKK.KKK_KKKKWKKKKKWKWKKKK__KKKK.KK
K..K.KKKK.K..K.WK.........K........K...K........................

And the "stuck" processes have very high SS:

-1 10328 0/4230/4251 _  39.99 98 12052 0.0 23.65 23.65  crtntx1-ar2-169-156.crtntx1.dsl xxxx.xxxx.net GET /perftest/text1.htm?1005325141989 HTTP/1.1 

The server will eventually refuse connections when it hits max clients.  I've read through similar reports w/Apache on Solaris and Linux, but found no resolution.
>How-To-Repeat:
Unsure.  I can repeat here by just letting the server run for a few hours.  This problem also happened in version 1.3.19 and I compiled 1.3.22 to hopefully fix it but to no avail.
>Fix:
No.  I've seen quick-fix patches suggesting sending USR2 instead of ALRM, but in my case, not even USR2 kills the process.  The only way I know to fix it is to turn keepalives off, or periodically restart the server, which both impact performance or stability, but at this point, is better than the server refusing connections.
>Release-Note:
>Audit-Trail:
>Unformatted:
 [In order for any reply to be added to the PR database, you need]
 [to include <ap...@Apache.Org> in the Cc line and make sure the]
 [subject line starts with the report component and number, with ]
 [or without any 'Re:' prefixes (such as "general/1098:" or      ]
 ["Re: general/1098:").  If the subject doesn't match this       ]
 [pattern, your message will be misfiled and ignored.  The       ]
 ["apbugs" address is not added to the Cc line of messages from  ]
 [the database automatically because of the potential for mail   ]
 [loops.  If you do not include this Cc, your reply may be ig-   ]
 [nored unless you are responding to an explicit request from a  ]
 [developer.  Reply only with text; DO NOT SEND ATTACHMENTS!     ]