[asterisk-dev] Suspected deadlocks in Asterisk 1.8 under heavy load

[Bryant Zimmerman]

----------------------------------------

From: "Paul Belanger" <pabelanger at digium.com>

Sent: Wednesday, August 17, 2011 10:22 AM

To: asterisk-dev at lists.digium.com

Subject: Re: [asterisk-dev] Suspected deadlocks in Asterisk 1.8 under heavy 
load


On 11-08-16 10:32 PM, Kevin P. Fleming wrote:

> On 08/16/2011 09:27 PM, Alistair Cunningham wrote:

>> On 17/08/11 12:23, Kevin P. Fleming wrote:

>>> Matt Nicholson committed a change to the 1.8, 10 and trunk branches

>>> today to solve a significant performance issue caused by the change to

>>> chan_sip to return the SIP hangup cause to the 'master' channel. His

>>> change made that behavior optional, even though it was already 
released

>>> in 1.8, because of the performance impact it has. We had another

>>> customer report a similar set of symptoms.

>>>

>>> If possible, it would be most helpful if you could try that patch on 
one

>>> of your affected systems before you downgrade it. I can understand if

>>> your customer is not willing to let you try that, though :-)

>>

>> Kevin,

>>

>> Thank you for this. When you say "optional", are any configuration

>> settings needed to disable it?

>

> Yes, in Asterisk 1.8 you'll need to set 'storesipcause' to 'off', since

> the default is 'on' to preserve the existing behavior. In Asterisk 10

> and later, the default will be 'off'.

>

> If we have real-world testing that shows that changing the default to

> 'off' will resolve issues such as yours, we'll consider even changing

> the default to 'off' for the Asterisk 1.8.6 release. It'd be an unusual

> step to take, but unless there is sufficient community demand for this

> feature to be enabled by default, the performance problems it causes are

> not acceptable for users who don't care about the feature.

>

I believe this is something we should look at doing, even thought it 

might be against policy. But like you said, this requires people to 

test 1.8.6.0-rc2 (assuming we merge the patch into it). Another option 

we discussed on #asterisk-dev is to override the settings in the samples 

configs, so new installations benefit from the performance change.


Ideally, sites still using 1.4/1.6.* with large volumes of calls would 

be the best testers to confirm the performance impact of enabling and 

disabling 'storesipcause' setting.

_________________


What is really lost by setting 'storesipcause' to off? As I understand it 
it does not effect global variables? What was the usage case for the 
feature in 1.8? How would this effect things long term?


Thanks

Bryant

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.digium.com/pipermail/asterisk-dev/attachments/20110817/6e244db3/attachment.htm>