Stipe, Thanks for the reply.
SIGTERM or SIGINT would in fact shutdown kannel? That's great, so I'll switch from SIGKILL then (ops!) :) I usually send an administrative shutdown from the web interface, before shutting down, but the corruption persists. BTW, it's intermitent, last night it restarted without problems. I experienced the same problem on two different setups and platforms (both Linux, one FC2 and one Centos4.4) Regarding this:
> 2. It would be _way_ cool to get the message ID from the offending > message on the log file, so it can be deleted. ???
I've meant to say that when kannel PANICs and die, it just says "PANIC: Not handled sms-type within store". If, for instance, the message could be "PANIC: Not handled sms-type within store. Message ID: xx/whatever-message-id", I could just delete the message from store and try again. If I'd like to change the behaviour from PANIC to WARNING, where should I start looking?. An alternate approach could be to move the failing messages to a "quarantine" folder, so they can be checked later. I could even setup a cron job to check on that folder, so I can be notified. That way, I can rest assured that kannel will keep on going no matter if one or two messages get corrupted. Regards, Alejandro On 2/4/07, Stipe Tolj <[EMAIL PROTECTED]> wrote:
Alejandro Guerrieri wrote: > Dear List, > > I've recompiled Kannel a couple of days ago to latest CVS, 20070201. > I've chose to use the new spool directory to store messages. > > Since then, I've experienced a strange problem when restarting kannel: > sometimes, the store seems to get corrupted, so I get a "PANIC: Not > handled sms-type within store" and Kannel stops. > > I've experienced a similar problem with an older install (20060830 or > near) on another server using the good ol' store file (sometimes the > store gets corrupted and I have to manually delete the store file to > get it going again) so I'm not sure if it's really related with the > kind of store I'm using but maybe with kannel not properly closing > files when it's shutting down > > The log file didn't show too much information. Even with the spool > directory, there's no way to determine what messages are causing the > PANIC. > > I've inspected the spool directory and found most messages sizing 366 > bytes, though many were smaller (294 and 296 bytes generally). I've > tried deleting some of them to no avail. I've finally gave up and > remove all folders from my store spool folder and Kannel started again > with no problems. > > So, my questions/suggestions are: > > 1. Is there anything I might be doing wrong to cause the message > corruption during restarts? What's the "proper" way to kill kannel and > start it again? at least SIGTERM or SIGINT should be used and no SIGKILL signal. So this means $ kill <bearerbox-pid> is ok and $ kill -9 <bearerbox-pid> is NOT ok. But I can't think of this as a "reason" for the corruptions. > 2. It would be _way_ cool to get the message ID from the offending > message on the log file, so it can be deleted. ??? > 3. What's the real meaning of "Not handled sms-type"? Obviously a non recoverable error while unpacking/interpreting the msg from the store space. > 4. Is there any way to just ignore/skip this message(s) and keep on going? Hm, Kannel would PANIC here obviously... We "may" discard any corrupted message and continue. The negative side here: you wouldn't have noticed that something/somewhat is corrupting your store file (/dirs). So a PANIC is always a serious signal to the user. Stipe ------------------------------------------------------------------- Kölner Landstrasse 419 40589 Düsseldorf, NRW, Germany tolj.org system architecture Kannel Software Foundation (KSF) http://www.tolj.org/ http://www.kannel.org/ mailto:st_{at}_tolj.org mailto:stolj_{at}_kannel.org -------------------------------------------------------------------
-- Alejandro Guerrieri Magicom http://www.magicom-bcn.net/ LinkedIn: http://www.linkedin.com/in/aguerrieri