APBs can run for a very short time ~30 seconds to a very long time ~30 minutes. The way we gather bind credentials has a timeout of 5 minutes. We need to allow for APBs to have all the time they needs to run.
Possible solutions:
1) Poll the APB and wait for an error or completion
Remove the broker timeout entirely and expect that if an APB hasn't exited with an error or hasn't completed bind-init then work is still being done.
An issue with this approach is that if a task hangs in a playbook, we have no timeout for it and the broker will forever look for the bind credentials.
2) Interrupt driven I/O
Use entrypoint.sh to signal the broker that the playbook has completed.
This could be a simple http request to the broker letting it know that now is a good time to gather credentials. In this scenario, there is no timeout on the APB and the Broker doesn't waste resources looking for credentials.
APBs can run for a very short time ~30 seconds to a very long time ~30 minutes. The way we gather bind credentials has a timeout of 5 minutes. We need to allow for APBs to have all the time they needs to run.
Possible solutions: 1) Poll the APB and wait for an error or completion
2) Interrupt driven I/O