[asterisk-biz] Recent REMOTE CRASH BUG

[Julian Lyndon-Smith]

Ming Yong wrote:
> Guys,
> I believe everybody agrees that Asterisk v1.4 is by no means production 
> ready. 

I disagree. We use it here, and have since the early betas from July 
last year.

I believe almost everyone who runs production level systems are
> still at v1.2.15, which is great by the way. 

We're running 1.4 svn in production.

Kevin has a point in that
> for the amount of features added in Asterisk v1.4. The team at Digium 
> has done a good job in general however, I have to agree with Matt that 
> given the immense financial returns digium has gotten from the Asterisk 
> project. There is definitely room for improvement on the "open-ness" and 
> stability of asterisk.

It is very stable for us. The only time it's really broken was when we 
were playing with esoteric stuff (jabber with tls). Since a month ago 
we've had maybe 1 crash that I didn't cause (I know that module reload 
app_queue.so has a bad effect if you have active calls in a queue at the 
time, and if my odbc database goes down, then the phone system goes 
down). If you know what can cause a problem, report it, and don't do it 
until it's fixed ;)

Our setup and config is

1 server, 2GB ram dell 2850 running centos 4.3
20+ queues, with 80 agents logged into multiple queues.
150 registered phones
All agents communicate with asterisk via jabber (and asterisk 
communicates with the agents via jabber). 100,000+ messages per day
All cdr records are into odbc
voicemail is odbc realtime
queues are odbc realtime
queue members are odbc realtime
5000+ calls per day

we are very happy with * and digium. Thanks guys !

Julian.

> 
> Ming
>> Matt wrote:
>>  
>>> However, we will not be purchasing any more Digium cards until Digium
>>> realizes that PCI is an extremely old standard.  As it stands right now,
>>> and someone may correct me if I'm wrong, all Digium cards are PCI-X, and
>>> hence have issues with sharing IRQs.    Unfortunately, as you will see
>>> if you do a search on the mailing list, most new motherboards insist on
>>> sharing IRQs with things... this causes audio issues.       
>>
>> If you believe PCI (or PCI-X) has anything at all to do with interrupt
>> sharing issues, you are mistaken. If you believe that PCI Express has
>> anything to do with solving interrupt sharing issues, again you are
>> mistaken.
>>
>> PCI, PCI-X and PCI Express all support the same interrupt handling
>> methodologies; PCI Express just provides a different signaling
>> mechanism. It is up to the system's chipset designers to decide how
>> those interrupts are going to be handled (XT-PIC, APIC, etc.). It would
>> certainly be possible for there to be a PCI Express system that did not
>> use an APIC for IRQ routing, but you'll never see one because PCI
>> Express systems are brand-new and all brand-new chipsets have APICs.
>>
>>  
>>> Until Digium gets with 2007, we will be purchasing Sangoma PCI-express
>>> cards.   I hate to do it, because like I said I want to support Digium,
>>> but their hardware is not staying with the times.
>>>     
>>
>> I won't disagree that our lack of PCI Express support is something we'd
>> like to solve, but IRQ sharing issues will not be solved by just
>> switching to PCI Express (which in most cases involves nothing more than
>> sticking a PCI Express -to- PCI-X bridge chip on the board). We have
>> already made some significant driver improvements in the last couple of
>> months in this area, and our customers (and support department) are now
>> reporting that in nearly every case of a suspected 'IRQ sharing' problem
>> updating to the latest Zaptel 1.2/1.4 release cures the problem (note
>> that I am not in support nor do I deal with these issues directly... I
>> am only reporting what I have been told).
>> _______________________________________________
>> --Bandwidth and Colocation provided by Easynews.com --
>>
>> asterisk-biz mailing list
>> To UNSUBSCRIBE or update options visit:
>>    http://lists.digium.com/mailman/listinfo/asterisk-biz
>>
>>   
> 
>