Opened 19 years ago

Closed 17 years ago

#1038 closed defect (worksforme)

FCGI on Windows does not kill mapserv processes

Reported by: banders@… Owned by: warmerdam
Priority: high Milestone:
Component: MapServer C Library Version: 4.3
Severity: normal Keywords:
Cc:

Description (last modified by hobu)

In some cases Apache/mod_fastcgi do not destroy all of my mapserver processes
when Apache is stopped.  Processes seem to be destroyed properly when there are
only a couple of them, but when there are 8-10 mapserv processes they are not
destroyed when apache stops.  This leaves orphened processes that may be holding
database connections.

My fast cgi config in httpd.conf:

FastCgiConfig -initial-env PROJ_LIB=c:/proj/nad -appConnTimeout 60 -idle-timeout
60 -init-start-delay 1 -minProcesses 2 -maxClassProcesses 10 -startDelay 5 

I have tested with different mapfiles.  I have tested with shapefiles and with
SDE data, and I consistently encounter the problem.  You should be able to
duplicate this with any simple mapfile.  I use JMeter to make 8-10 concurrent
getMap requests, which encourages the fcgi server to start several mapserv
processes.

Attachments (1)

httpd.conf (34.8 KB ) - added by banders@… 19 years ago.
my httpd.conf

Download all attachments as: .zip

Change History (3)

comment:1 by fwarmerdam, 19 years ago

Status: newassigned
Brock,

I tried a couple of days ago to reproduce this and failed.  I was able to 
force 10 mapserv fastcgi servers to run but shutting down apache terminated
them all.  

I am wondering if you might have any directives in your http.conf file that
relate to how long it is allowed to take to shutdown?  I don't see any control
of this in my conf file though.


by banders@…, 19 years ago

Attachment: httpd.conf added

my httpd.conf

comment:2 by hobu, 17 years ago

Description: modified (diff)
Resolution: worksforme
Status: assignedclosed

Brock,

Please re-open this if it is still an issue. The SDE driver has seen a number of updates (hopefully for the better) since this issue was reported, especially in cases where it might have been doing nasty things like corrupting the heap. IIRC, things were behaving quite well at the end of our SDE join development work.

Howard

Note: See TracTickets for help on using tickets.