Re: Message retry and backoff

hi

sounds good.. Under what conditions do you retry?

Nisan
At 04:13 PM 10/22/02 +0200, Oded Arbel wrote:
>Hi list.
>
>we added a patch to Kannel to enable message retry and backoff. it was 
>quite simple really, but I'm pretty sure it can mess things up, so I would 
>like to get suggestions as to how to properly handle this . it goes 
>something like this :
>- we add to the message structure to integers, retry and delay.
>- whenever smscconn_send() is called on a message, and before it stores 
>that message in the driver's queue it will increment the retry field by 
>one, and add to the delay field a configured number of seconds (we set it 
>currently to 60) times the number of retries (this achieving the backoff 
>effect).
>- sms_router(), before calling smsc_rout() we do this :
><snip>
>         if (msg->sms.retry > BB_SMSCCONN_MAX_RETRY) {
>                 warning(2, "Message has too many retries - consider it a 
> failure");
>                 bb_smscconn_send_failed(NULL, msg, 
> SMSCCONN_FAILED_DISCARDED);
>                 ret = 0;
>                 continue;
>         }
>
>         if (msg->sms.delay + msg->sms.time > time(NULL)) {
>                 debug("bb.sms", 0, "Message is delayed - will send later, 
> message retries %d", msg->sms.retry);
>                 list_produce(outgoing_sms, msg);
>                 ret = 0;
>                 continue;
>         }
></snip>
>
>the ret=0 emmulates the effect of smsc_rout() returning after not founding 
>any available drivers.
>
>comments please ?
>
>--
>Oded Arbel
>m-Wise mobile solutions
><mailto:[email protected]>[email protected]
>
>+972-9-9581711 (116)
>+972-67-340014
>
>::..
>-- Pulchritude possesses solely cutaneous profundity.