Channels and calls lost in SQLite

2 Sep 2010

      Hello guys,

I have a problem that started to happen after I upgraded to git head last
night. No configuration changes were made and this is the intriguing part.

I have SQLite being mounted on tmpfs on busy Dell 1950s. After some time on
the busy moments of the day, FS starts to not delete channels/calls from
SQLite and leaves very little trace for me to track.

I grep'ed the logs for SQL errors and found nothing.. The only part of the
log that I could spot and that called my attention was this line that
repeated several times during the periods where the problem occurs:

switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_WARNING, "queued event at a
lower priority %d/%d!\n", index, was);

11:23:13.977874 [WARNING] switch_event.c:1353 queued event at a lower
priority 1/0!

Which is printed from the following code:

                   for (index = (*event)->priority; index < 3; index++) {
                        int was = (*event)->priority;
if (switch_queue_trypush(EVENT_QUEUE[index], *event) ==
SWITCH_STATUS_SUCCESS) {
if (index != was) {
switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_WARNING, "queued event at a
lower priority %d/%d!\n", index, was);
}
goto end;
}
...
From what I could understand on the code, seems like there are 3 event
priority queues and that message occurs when queue 0 is full. How exactly
that relates to the problem, I am not sure so I see myself here lost in
translation. Could the core_event_handler not receive some messages
therefore not execute the sql statements? If that's the case, maybe I lost
some log line telling me that we were loosing events?
I hope someone can shed some light.

Thank you,
João Mesquita

João Mesquita

Moises Silva

Michael Collins

Jeff Lenk

Anthony Minessale

Sergey Okhapkin

João Mesquita

Anthony Minessale

tags

participants (6)