[chef] Re: Why don't work search and how to search in run_list (for example with knife search)?


Chronological Thread 
  • From: Graham Christensen < >
  • To:
  • Subject: [chef] Re: Why don't work search and how to search in run_list (for example with knife search)?
  • Date: Mon, 19 Aug 2013 08:24:00 -0500

Hello,

This is by design - if search were to return nodes which haven't run
the recipe yet, you could very easily imagine a situation where other
services start failing, because it can't access a database server, or
a load balancer is pointing at nodes which don't even have the service
running yet.

There may be a way to force the issue and search for nodes who don't
have the recipe run yet, however usually when I hear of people wanting
to do this, they are looking for immediate configuration to be
performed, instead of assuming eventual infrastructure consistency.
This means running the server's chef-client, running it on the client,
and then re-running it on the server. Often this is accomplished by
having chef-client be scheduled every 5 minutes or every hour.

Consider it  pessimistic view:

Either it:
1. assumes the chef-client will prepare the node for service very
soon, and potentially causes failures in your infrastructure
2. it makes no assumption about the state of the node, and simply
recognizes that it simply isn't ready for service yet


Does this make sense?

--
Graham Christensen





Archive powered by MHonArc 2.6.16.

§