Hi list,
when having a high IO scenario, i.e. a lot of DLRs coming in from the
upstream SMSC side, and the smsbox connection vanished, then the thread
gw/bb_box.c:sms_to_smsboxes() is most likely in the gwthread_sleep()
loop captured. If we SIGTERM bearerbox it will iterate one more time,
since the bb_status hasn't reflected the fact that we're in shutdown now
and we need to wait 1m until the bearerbox goes down.
This is unnecessary IMO, since the MO/DLRs are persistent in the store,
and we're not accepting any smsbox connection anyway when in shutdown
mode. So we can make sure we wakeup-all in the main() scope and test for
the corresponding bb_status in the while loop of the thread.
Please find the corresponding patch attached. If no objections arise,
I'll commit the patch during the end of the week.
Stipe
--
-------------------------------------------------------------------
Kölner Landstrasse 419
40589 Düsseldorf, NRW, Germany
Kannel Foundation tolj.org system architecture
http://www.kannel.org/ http://www.tolj.org/
mailto:stolj_{at}_kannel.org mailto:st_{at}_tolj.org
-------------------------------------------------------------------
Index: gw/bb_boxc.c
===================================================================
--- gw/bb_boxc.c (revision 5138)
+++ gw/bb_boxc.c (working copy)
@@ -1681,7 +1681,7 @@
newmsg = startmsg = msg = NULL;
- while(bb_status != BB_DEAD) {
+ while (bb_status != BB_SHUTDOWN && bb_status != BB_DEAD) {
if (newmsg == startmsg) {
/* check if we are in shutdown phase */
Index: gw/bearerbox.c
===================================================================
--- gw/bearerbox.c (revision 5138)
+++ gw/bearerbox.c (working copy)
@@ -789,6 +789,9 @@
/* call shutdown */
bb_shutdown();
+ /* wake up any sleeping threads */
+ gwthread_wakeup_all();
+
/* wait until flow threads exit */
while (gwlist_consume(flow_threads) != NULL)
;
