[asterisk-bugs] [Asterisk 0012127]: Using state_interface and Local channels allows several simultenous calls to be sent to agent

[noreply at bugs.digium.com]

A NOTE has been added to this issue. 
====================================================================== 
http://bugs.digium.com/view.php?id=12127 
====================================================================== 
Reported By:                atis
Assigned To:                putnopvut
====================================================================== 
Project:                    Asterisk
Issue ID:                   12127
Category:                   Applications/app_queue
Reproducibility:            sometimes
Severity:                   minor
Priority:                   normal
Status:                     assigned
Asterisk Version:           SVN 
SVN Branch (only for SVN checkouts, not tarball releases):  trunk 
SVN Revision (number only!): 103809 
Disclaimer on File?:        N/A 
Request Review:              
====================================================================== 
Date Submitted:             03-03-2008 12:26 CST
Last Modified:              03-11-2008 13:17 CDT
====================================================================== 
Summary:                    Using state_interface and Local channels allows
several simultenous calls to be sent to agent
Description: 
I've seen this situation on 8-CPU 3GHz Xeon, that agent gets 4 queue calls
simultaneously. 

For now i've been able to avoid this by checking/setting group at
beginning of Local channel, setting call-limit=1 for SIP peers (only for
those who shouldn't receive more than 1 call simultaneously) and checking
group upon unsuccessful Dial from Local channel. Those 3 fixes together are
giving some good results for one most-problematic queue, however that's not
a permanent and stable solution. If queue is creating several channels at
the same time, on slower but multi-cored systems this could lead to race
condition between GROUP_COUNT and Set(GROUP()) in two parallel channels,
and call-limit=1 also can't always be set.

I'm not sure if this can be implemented in good way, but I think the most
appropriate solution for this would be that Queue is setting device state
of Local channel, and other queue(s) (if member is in multiple queues)
receives event from first queue's devicestate change.

My system is Asterisk 1.4.14 with backported state_interface from r103809
====================================================================== 

---------------------------------------------------------------------- 
 putnopvut - 03-11-08 13:17  
---------------------------------------------------------------------- 
I did a bit of looking and testing, and it seems like the
update_dial_status function should be used in the case that an ast_request
fails. Think about the following scenario: You have a SIP device as a
member in the queue, then you unload chan_sip.so. If you do a "queue show"
at this point, then the SIP device will still show up as "Not in Use" when
it should be "Invalid." The first time the channel is requested, the
failure will cause the status to be updated correctly to "Invalid." If
chan_sip is loaded later, then the device state thread will take care of
returning the device back to "Not in Use."

So, here are the changes I'm going to make.

1. Remove all but the one call to update_dial_status in the case where
ast_request fails.

2. Since this indicates that some sort of failure occurred, we should
never set the member's status to "unknown." I will remove the part of
update_dial_status that sets the member's status to "unknown." 

Issue History 
Date Modified   Username       Field                    Change               
====================================================================== 
03-11-08 13:17  putnopvut      Note Added: 0083736                          
======================================================================