The broker at present will log warnings about overrunning timer tasks
and late wakeups. Not surprisingly, a common question from users is
whether they should be concerned with these statements and how they
prevent them.
My belief is that in general they can be ignored. The information in the
warnings is only of interest if they are experiencing other problems
that may be connected to the precision of timer tasks. The most obvious
issue - for which I believe these warnings were introduced - is where
clients are timing out due to delays to broker heartbeats.
Tasks that require a degree of precision (e.g. heartbeats) to function
correctly imply constraints on other tasks driven by the same timer
thread. If this is a concern these constraints should be made explicit.
We can then either have different timer instances with different
constraints, or separate the processing of (longer running) tasks from
their activation.
Going back to the logs however, I think these warnings should be demoted
to informational statements (at most). It is much easier to turn them on
if needed (--log-enable info+:Timer) than it is to turn them off if not.
I know this has been raised before... any objections to making this change?
---------------------------------------------------------------------
Apache Qpid - AMQP Messaging Implementation
Project: http://qpid.apache.org
Use/Interact: mailto:dev-subscr...@qpid.apache.org
