search

autodesk-cloud / ochopod

Your friendly orchestration overlay over Mesos, K8S and more !
http://autodesk-cloud.github.io/ochopod/
Apache License 2.0
122 stars 20 forks source link

Ability to leave managed process running #14

Closed stphung closed 9 years ago

stphung commented 9 years ago

I tried overriding the tear_down implementation to be empty such as:

def tear_down(self, process):
  pass

This results in the ochopod configuration blocking however and still waiting for the process to be eventually terminated by the system after about 1 minute. It would be awesome if I could leave the existing process running so that I can have more control on the next configuration run.

My use case is zero downtime haproxy configuration reloads. I basically need to have the process running so I can start a new haproxy process which can signal to the currently running process.

opaugam commented 9 years ago

ok - simply define 'soft=True' in your pod.py strategy and ochopod will not attempt to force a SIGKILL should your tear_down() implementation be a no-op.