Dealing with Restarting Mesos
solarkennedy opened this issue · comments
I have my own mesos profile that handles the configuration of this module to meet my needs.
Part of that profile includes logic that handles restarting mesos when configs change.
I know that is a controversial thing, not everybody believes that puppet should be doing that, but I do.
What is your take on this, and would you accept a patch that restarted mesos? (if not, would you accept it if it was toggle-able and defaulted to false?)
Hey, I don't have a problem with restarting Mesos services when config changes. In fact, that is what many Puppet modules do. It would be good to test that this features does not cause restarts when nothing was changed.
Cool. I don't know if there is really a way to test that part in unit tests, but I can tell you from the code I have at work, it does not continuously try to restart the service.
I'll take parts of the code that are legit and do a PR.
I have other "hacks" that do things like remove stale pid files and clear out the slave metadata when the slave can't start, mostly because I always want puppet to be able to fix slaves that can't get going after a bad config push. But these things I think are not really appropriate for the module. The module, I think, should assume you are passing legit config hashes and junk.
I've added in b728737 notification to services when some property changes. Does it solve some of problems you were dealing with?
Ah! I does. Thank you.
I'm going to close for now. The other service restarts take place when the package changes in the postinst scripts when the package updates, so thats cool.