[rt-users] RT creates CLOSE_WAITS

[Bill Cole]

On 28 Oct 2015, at 4:07, Asanka Gunasekera wrote:

> Hi all I have been struggling with this CLOSE_WAITS.
>
> I have an Request Tracker installed on CEntOS 6.7 (64 bit) RT 
> rt-4.2.12 and Apache httpd-2.2.31. I can log in and all, but after a 
> while I am seeing CLOSE_WAIT socket are created for each selection 
> that I make on the dashboard. Eventually the system goes unresponsive.
>
> root at iplrt logs]# netstat -nalp | grep -i wait
> tcp 1 0 192.168.11.254:80 192.168.11.118:57076 CLOSE_WAIT 3176/httpd
> tcp 1 0 192.168.11.254:80 192.168.11.118:57083 CLOSE_WAIT 3177/httpd
> tcp 1 0 192.168.11.254:80 192.168.11.118:57082 CLOSE_WAIT 3185/httpd
> tcp 1 0 192.168.11.254:80 192.168.11.118:57084 CLOSE_WAIT 3365/httpd
> [root at iplrt logs]#
>
> I don't see any errors in the logs! I have downgraded both RT and 
> Apache with no luck and I am STUCK no ware to goooooo
>
> Can some one tell me how to go about diagnosing this

What interface are you using to run RT via Apache? (Possible answers: 
mod_fastcgi, mod_fcgid, mod_perl 1.x, mod_perl 2.x)

That is important because a CLOSE_WAIT state indicates that the OS is 
waiting on a local process (in this case multiple Apache httpd 
processes) to recognize that the socket has been closed on the other end 
and call close() itself. The Apache core doesn't have that problem much, 
so the issue is most likely lurking in whatever module is loaded into 
the httpd process to handle running RT proper.

(my bet is that you're running a slightly broken mod_perl, but I'm 
biased by bad experiences...)