Creating and updating nodes "from file", any potential issues?


#1

Myself and M4rcu5 (Marcus Van Dam) are both considering creating nodes and
updating them using the “from file” rather than using the “runlist add”,
“run_list remove”, -E new_environment, etc. commands. This gives the
advantage that node configuration can be kept under git control.

Are there any potential issues with this approach that we should be aware
of?


#2

On Friday, November 25, 2011 at 2:04 AM, Bryan Berry wrote:

Myself and M4rcu5 (Marcus Van Dam) are both considering creating nodes and updating them using the “from file” rather than using the “runlist add”, “run_list remove”, -E new_environment, etc. commands. This gives the advantage that node configuration can be kept under git control.

Are there any potential issues with this approach that we should be aware of?
If you’re using Hosted Chef, you’ll need to create the API Client for the node first, and use that to create the node. In Hosted Chef, clients can only update nodes that they “own” (that is, nodes that they created), so creating the node using an admin account will not give the client the required privileges.

If you’re using the open source server, it should work fine.


Dan DeLeo


#3

On Nov 25, 2011 1:04 AM, “Bryan Berry” bryan.berry@gmail.com wrote:

Are there any potential issues with this approach that we should be aware
of?

It sounds like you want to treat nodes as being individually special, which
best practice usually discourages. At Opscode we have runbook instructions
on how to create each type of server and treat the existing ones as
disposable resources.

I would check that you aren’t overloading node attributes which would
increase the difficulty to duplicate that node and build arcane knowledge
into an existing nodes configuration.

Bryan


#4

btm, good points, I will have to think about this

On Tue, Nov 29, 2011 at 9:35 AM, Bryan McLellan btm@loftninjas.org wrote:

On Nov 25, 2011 1:04 AM, “Bryan Berry” bryan.berry@gmail.com wrote:

Are there any potential issues with this approach that we should be
aware of?

It sounds like you want to treat nodes as being individually special,
which best practice usually discourages. At Opscode we have runbook
instructions on how to create each type of server and treat the existing
ones as disposable resources.

I would check that you aren’t overloading node attributes which would
increase the difficulty to duplicate that node and build arcane knowledge
into an existing nodes configuration.

Bryan