| 5.5.1 on Linux We have a host system with many containers, all managed by Puppet. When the hosts boots or all containers need to be restarted for some reason, all Puppet agents start at the same time. We do not want to use splay since that would affect Puppet restarts we perform manually after changing configuration. It would be useful if Puppet had some sort of setting, similar to pre_run_command and post_run_command to allow an external script to decide whether a Puppet run should start now. I imagine this to work as a polling mechanism, where Puppet polls an external script repeatedly for permission to run now after its own runinterval has expired. I do not want to implement this via pre_run_command since i do not think this is designed for arbitrary wait intervals. This could solve our problem by allowing us to implement our own worker pool logic in e.g. a shared directory, where a Puppet agent needs to lock one of a limited number of slots before it can start its run. We could then free that slot again in post_run_command. |
