One way to do this is to change the docker container to run the server using supervisor,
then add this to supervisor.conf:

  [eventlistener:roundup_mail]
  ## poll for mail to the tracker every minute.
  command=./mailgw_poller demo mailbox mbox
  events=TICK_60    # poll every minute
  autorestart=true  # always run regardless of error status of last run.
  autostart=false

then create the poller script:


#! /bin/sh
# used by eventlistener protocol transition to READY state
printf "READY\n"
# wait for event to trigger next step
read header
# import mbox on trigger
python roundup/scripts/roundup_mailgw.py "$@"
# report ok and exit to start next cycle
printf "RESULT 2\nOK"


A few caveats, this is an abuse of supervisord, but it worked.

If roundup_mailgw takes > 1 minute supervisord will queue the events for this service.
Can cause a backlog.

Need to: use roundup-mailgw in production (I was testing from a dev tree)

         create a watchdog to kill roundup-mailgw if it takes > 1 minute

roundup-mailgw reports errors to stderr so it is captured by
supervisord. Exit status for roundup-mailgw is 1 in most error
cases.