Atomicity of updates on the Puppet master

[Mitchell Hashimoto]

Hello,

I'm currently setting up some scripts that will pull in new data for Puppet master (modules, manifests, hiera data, etc.) from a Git repository. It works great!

However, I'm concerned about atomicity of the whole thing. Does the Puppet master cache the modules/manifests when it starts? If so, do I need to restart the Puppet master on each update? Or can I send it a SIGHUP? etc.

Best,

Mitchell

[Trevor Vaughan]

If you update any server side functions or native types then I would
restart the PM or httpd if you're running Passenger.

The PM is *supposed* to note all new updates but I've seen cases where
this does not always happen.

For anything that's a pure manifest, I have not had an issue.

Trevor

> --
> You received this message because you are subscribed to the Google Groups
> "Puppet Users" group.
> To post to this group, send email to puppet...@googlegroups.com.
> To unsubscribe from this group, send email to
> puppet-users...@googlegroups.com.
> For more options, visit this group at
> http://groups.google.com/group/puppet-users?hl=en.



--
Trevor Vaughan
Vice President, Onyx Point, Inc
(410) 541-6699
tvau...@onyxpoint.com

-- This account not approved for unencrypted proprietary information --

[Mark Roggenkamp]

Even if it cached individual files wouldn't it still be possible for it to grab two related files, one just updated and one not yet updated to be used during a catalog compilation?

I think I'm going to treat modules/manifests together as an artifact (tarball probably) and put a new (probably versioned) artifact in another folder and symlink /current to it and use mv -T as described here: http://blog.moertel.com/articles/2005/08/22/how-to-change-symlinks-atomically.

Sound reasonable?

Thx

Mark