[asterisk-dev] CEL backend locking Asterisk event processing

[Kirill Katsnelson]

Here's what I am seeing. A CEL backend (MS SQL server) was taking long 
time to insert records because of some problem not really important 
here. What's important was that it could not cope up with the stream of 
CEL events.

At that time, I have seen many calls that went out of bridge, but still 
sitting in channel Up state. For some of these it took 1.5 HOURS to 
complete, about as long as the backlog of CEL events was. As if a long 
CEL queue was locking up the core_event_dispatcher task processor.

Is that an explanation? What can/should be done about that? Should CEL 
get its own taskprocessor, separate from  core_event_dispatcher? There 
would be an issue with too many event piling up, but that can be dealt 
with using thresholds on the size of the CEL event queue.

Am I making any sense?

  -kkm