Hi,
We are using the http queueing facility within the smsbox to handle the
cases where there might be outages of our internal systems. On testing
recently following a loarge outage of one system, I have found some
strange behaviour.
We had approx 400 messages in the retry queue due to the outage. each
message was logged in the smsbox.log file with an ERROR that the content
could not be fetched, however there was no indication that at all that
any retries were being attempted, nor how many messages were queued (we
don't run debug logging by default). I then found that the messages
were being resent, one at a time, every http-queue-delay interval, even
at the default of 10 seconds for the delay, that is one hour to resend
400 requests.
I did initially attempt to add some code to test / send every message in
the queue at the interval instead of one at a time, but my code
obviously had issues so given that I can resend the data myself by hand
quicker than kannel can resend it I settled with a simpler option of
updating our policy in the instance of a longer outage to restart the
smsbox and clear out the queued messages, then resend by hand.
Could I suggest the trivial patch of changing the DEBUG statements in
smsbox.c::http_queue_thread that indicate the queue length and the retry
action into INFO statements instead, giving some visibility into the
actions of the queue without having to resort to enabling debug logging
permanently.
--
Stuart Beck
-----------
Systems Administrator
m.Net Corporation
No one knows Mobile like m.Net
Tel. No. +61 8 8210 2049
Web: http://www.mnetcorporation.com
Try our new online mobile marketing system
www.campaignmobile.com
ADMA MMA Promotional Campaign of the Year, 2008 Winner .
BRW. FAST 100 Awards, 2008. Number 5.
Deloitte Technology Fast 50 Australia 2008 Winner.
