The producer uses post (async) to not block the Oracle DCN listener thread
comming from the JDBC driver so it can return asap.
This does not happen very often, like once a month where timely distances
may vary (no pattern), and indeed i see blacklist logs from Felix if i
remember correctly. The only solution on the net i found was setting the
timeout to 0, e.g. no timeout at all. This i find very troublesome, blocking
event-threads for a long time or even permanent, destabilizing the whole
node? So my timeout is set at 30 seconds which should be sufficient as the
JVM params i monitored are seldom at max levels.
On the consuming side the event is handed over ASAP to a ExcecuterService to
free up the event producing thread, which is comming from Hazelcast. If this
is blocked you will see the following without blacklisting logs, but also
stopping further eventing after release of the breakpoint:
Side question: is this consumer call synchronous comming from Hazelcast into
the OSGi EventAdmin?
So this seems to come out of nowhere from the producing node as i cannot
find any exceptions not system specific limitations, speculating on network
glitches? But then i should see more exceptions right?
-----
if ( you want ) { you can } else { you can't }
--
Sent from: http://karaf.922171.n3.nabble.com/Karaf-User-f930749.html
