Bug #127

fastcgi processes aren't cleaned upon sigkill

Added by Anonymous 1334 days ago. Updated 89 days ago.

Status:Invalid Start:
Priority:High Due date:
Assigned to:- % Done:

0%
Category:mod_fastcgi
Target version:1.4.9
Pending:

 Resolution:

invalid
Patch available:

Description

expected behavior:

in lighttpd 1.3.13, when start-stop-daemon is used to stop the lighttpd daemon, all fcgi processes spawned by the configfile method (internal spawning) are also closed as well.

current behavior:

in svn 12/05/2005 lighttpd, when start-stop-daemon is used to stop the lighttpd daemon, all internally spawned fcgi processes are left alive, without SIGKILL or other sent to them to ask them to shut down.

result:

in machines with limited memory, race condition ensues if fastcgi processes are assumed to be garbage collected and they are not, especially php-fastcgi processes that are allowed 16M of memory each.

-- infowolfe

History

08/18/2005 11:40 AM - jan

  • Status changed from New to Assigned

08/18/2005 11:44 AM - jan

  • Status changed from Assigned to Fixed
  • Resolution set to invalid

KILL is KILL, it can't be handled. Use TERM to terminate the process the nice way.

10/10/2008 06:51 PM - stbuehler

  • Status changed from Fixed to Invalid

Also available in: Atom PDF