fred wrote:
There is still a window of oportunity for the smsbox to service sendsms
requests with Sent! before shutting.
lol the "If you do kill bearerbox twice in a row"
why is that ?, very usually I need to issue 2 shutdown commands before it
seems to want to shutdown. (but not always)
I have implemented further checking of bb_status in various loops,
and created a shutdown admin msg to send to smsbox.
However, It is still not completely satisfactory yet.
Killing smsbox first may be a answer, but its manual work to
hunt out the processid, issue a kill command etc etc...
Fred
I'd sum up to this:
I agree that on a high-loaded system while doing a "normal" Kannel shutdown,
means signaling SIGTERM to bearerbox, and hence passing an admin msg from
bearerbox to smsbox requesting smsbox to terminate, there is a "opportunity
window" (while threads are still active) that the HTTP server module accepts
HTTP reuqests, they are processed and queued in the internal gwlist struct for
the upstream (bearerbox).
So, we either need to make that queue persistant, aka store-file, spool-dir
usage, or as a first punch, make sure we handle the internal thread shutdown
logic in smsbox that way that we garantee that
1) HTTP accept is stopped
2) queue towards bearerbox is emptied by bearerbox
3) shutdown of the rest
Means we need to garantee that any MTs that are currently in smsbox's scope are
passed to bearerbox at least before we say "good night".
Thanks Georg for point this out clearly ;) also to Fred for his forensical
analysis ;)
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
-------------------------------------------------------------------
