Re: Kannel store issue

Hi Donald,

which type of store you use? I think, you use file?
please try different store type and then see if it helps.

Thanks,
Alex

Am 02.08.2009 um 10:11 schrieb Donald Jackson:

> Hi Nikos,
>
> No its not dependant on queue size and is fairly unpredictable. I  
> see it does happen when the SMSC's have some sort of queues built/ 
> building, but not any specific size.
>
> From the store dump's I did all messages in store are MT-PUSH.
>
> Thanks again,
> Donald
>
> 2009/8/2 Nikos Balkanas <[email protected]>
> Hi Donald,
>
> A clarification in (2). You are thinking in terms of at driver. I am  
> thinking in terms of queue size. Is Q large and at a certain sixe  
> (i.e. 32768) when problem starts, or it could happen with a small Q  
> as well (i.e. 100). If it is depended on Q size, then it would  
> affect other SMScs as well, and will be at a different point in the  
> code.
>
> A last question. Both MO and MT SMS are affected in this state?
>
> BR,
> Nikos
> ----- Original Message -----
> From: Donald Jackson
> To: Nikos Balkanas
> Cc: kannel_dev_mailinglist
> Sent: Saturday, August 01, 2009 8:28 PM
> Subject: Re: Kannel store issue
>
> Hi Nikos,
>
> Just to answer your questions.
>
> 1) Yes, this is a bearerbox issue, restarts of smsbox / smppbox  
> alike do not effect the store (in this scenario)
> 2) At this stage unreproducible on demand as it is not one specific  
> modem giving the issue, the message(s) could be destined for any of  
> the given modems but the problem is there are so many modems (250)  
> and its on a remote site so its difficult to pin point, hence trying  
> to solve it 'heuristically'
> 3) No, the messages stuck in store are completely independent of  
> each other.
>
> My suspicion is there is some or other condition (possible race  
> condition?) in the AT driver which causes this scenario and some  
> messages to go 'unacked'. I hope I/we can find a solution as I'm  
> sure there are a number of us struggling with this problem in high  
> volume instances.
>
> I am continually looking but I'm thinking 'more eyes' will possibly  
> help quicker.
>
> Thanks,
> Donald
>
>
>
> 2009/8/1 Nikos Balkanas <[email protected]>
> Hi Donald,
>
> If you are inexperienced with the at driver, I am inexperienced with  
> SMS. However, I know a lot of kannel's code, and have my suspicions.  
> Since i do not have any SMSc connections, I hope it is possible to  
> recreate with full logs in your system. A couple of questions:
>
> 1) I don't doubt from what you say, that this is a bearerbox issue,  
> but to be sure, have you tried restarting smsbox only, while in such  
> state, to see if there is any effect?
> 2) From the logs, do you see SMS progressively delaying sent out  
> more and more, until they stop altogether and stay in queue? In  
> other words is it fairly reproducible, that this will happen after  
> so many messages, or more or less unpredictable?
> 3) Are there particular messages that are responsible for hogging  
> the queue, or could it be any message?
>
> Thanx,
> Nikos
> ----- Original Message -----
> From: Donald Jackson
> To: kannel_dev_mailinglist
> Sent: Saturday, August 01, 2009 4:50 PM
> Subject: Kannel store issue
>
> Hi everyone,
>
> I am moving this issue to the devel list to see if we can get any  
> more insight into it. It appears there are other users (from users  
> list) having the same issue with no 'user type' resolutions.
>
> I am running a Kannel installation (cvs-20090721) with roughly 250  
> AT (smsc_at.c) SMSC's. The problem I am experiencing is that after  
> some period of time it appears messages in the bearerbox get stuck  
> in the store. I have used store-status and Alex's store tools to  
> export the store and they all look like valid MT-PUSH messages (SMSC- 
> ID is routable, etc).
>
> The only way (seemlingly) to get these messages out of the store is  
> to restart the bearerbox, where they are pulled in and sent.
>
> My understanding of the store is that only messages which are ACK'd  
> on the SMSC [smsc_at.c / bb_smscconn_sent()] layer are removed from  
> store. I have not encountered this issue with SMPP binds, so I am  
> presuming this is related to the smsc_at.c driver. I am very  
> inexperienced with this so I would be very appreciative if anyone  
> with some experience with the smsc_at.c driver could possibly  
> provide some insight.
>
> I was previously running 1.4.3 with the same issue.
>
> Thanks in advance,
>
> -- 
> Donald Jackson
> http://www.ddj.co.za/
> donaldjster(a)gmail.com
>
>
>
> -- 
> Donald Jackson
> http://www.ddj.co.za/
> donaldjster(a)gmail.com
>
>
>
> -- 
> Donald Jackson
> http://www.ddj.co.za/
> donaldjster(a)gmail.com