initiate puppet run on agent from Puppet Master

asked 2013-06-20 10:47:47 -0600

rachelandrew

flag of United Kingdom of Great Britain and Northern Ireland

23 ●1 ●2 ●6 http://rachelandrew.co...

updated 2013-06-21 02:49:17 -0600

ramindk

4606 ●54 ●71 ●126

I currently have a Puppet Master and two agents, on the Puppet Master I am updating a json file and using that (via hiera) to get a list of running sites and deploy them to one of the agents.

If a site appears or is removed from the json, it is added or removed from the sites on the agent and this all works nicely if I run Puppet on the agent.

I expect to provision 20 or so sites per day (these are demos of our product) and it seems wrong to be constantly doing a puppet run on ... (more)

edit retag flag offensive close merge delete

add a comment

3 Answers

Sort by » oldest newest most voted

answered 2013-06-20 13:54:50 -0600

GregLarkin

4932 ●5 ●22 ●64 http://www.puppet.com/

You could reduce the run interval on that agent from the default 30 minutes to whatever you like.

Note that a value of "0" means "run continuously". I haven't experimented with that to see if there are any problems doing that, but maybe others who have can post here with their comments.

Here's another option for triggering puppet agent remotely using pssh.

edit flag offensive delete link

Comments

I'm a bit nervous about running Puppet continuously - although I may have no reason to be. Would be interested in other experiences. Triggering it using pssh would definitely be ...(more)

rachelandrew ( 2013-06-20 14:39:08 -0600 )edit

add a comment

answered 2013-06-21 01:39:29 -0600

ramindk

4606 ●54 ●71 ●126

updated 2013-06-21 01:52:19 -0600

Assuming potential demo-ees are hitting a button and data is being published into json that ends up in your Puppet master a few ideas present themselves.

As Greg mentioned, pssh to trigger the runs kicked off after an appropriate time later to avoid the race condition.
Mcollective might be a possibility in the near future if the Redis backend goes mainstream. Currently it's a fair amount of work to get the middleware required for Mcollective up and running on Debian based distros in my experience.
Any sort of async job system. Resque, delayed_job, gearman, etc. All of them will ...

(more)