When shell module plugins failed and exited,
confluent was failing to release the filehandle. This
would ultimately lead to unreasonable exhaustion of
filehandles.
If a bg sync thread is in progress, a key could be marked dirty before the desired change is actually made.
This causes the running thread to pick up the change to 'dirty' keys and save off the as-yet unchanged key.
Reorganize things to have the keys marked dirty only after they have been assigned so that an in-progress
commit to disk only picks up a key after the relevant values have changed.
Previously, the null tenant always synced to disk on every request in pracitice.
Correct that mistake to avoid commit to disk when no changes are being made.
ssl wrap prior to accept causes accept() to be too complicated to stay in the
persistent thread and makes key changes require restart to pickup. Call
the wrap_socket within the dedicated client thread so that it gets up to date
at the right time and picks up certificate changes in a timely fashion.
It turns out that eventlet.green.threading.Event() doesn't behave very efficiently in this context for whatever reason.
Use eventlet.event.Event() instead. It was not used before due to lack of timeout and clear, but that is overcome by
disposing of it rather than reusing and using with eventlet.Timeout() to add timeout to wait that doesn't have built in timeout.
If wanting to run as non-root, mkdir -p /var/run/confluent /var/log/confluent /etc/confluent
and chown those to be owned by confluent user. That is probably path for deb and rpm packaging.
