Closed jaypipes closed 11 years ago
So the spec start
will automatically destroy the nodes after the spec has completed. I had thought I had it set to auto delete after a successful run, but it looks like it just does by default all the time. I'll update that, since for me the proper behavior is to leave the cluster for inspection on failure, but to clean up if everything was successful. You can turn off the auto destruction using --no-auto-destroy
The --irl
is a some what different thing. The intention for it is to break out of the vagabond eco-system and hit nodes running "in real life". This means it will use the configured chef-server in the knife.rb instead of redirecting internally, and match nodes based on the defined run-lists in the cluster, and then spec them. The intention is to allow running specs against production itself to ensure the assumptions made within the specs actually pass in the real world.
On a side note, the Layout
file is still pretty clunky, so thanks for enduring it :) It will be getting merged into the Vagabondfile in the near term future.
Hi Chris,
This particular patch fixes an issue in the documentation where the "start" is missing from the spec command :) If I don't use "start", I get failures:
jpipes@uberbox:~/repos/att-cloud/cookbook-infra-db$ bundle exec vagabond spec metering_db
Could not find command "metering_db".
The README just said that you needed to issue vagabond spec
Best, -jay
Ah, sorry, I completely read over the fact that you were using --irl
for its intended purpose last night. Awesome!
I had some local changes already staged so I'm just cherry-picking this out.
Thanks for the PR :+1:
Docs were incorrectly calling vagabond spec my_cluster instead of vagabond spec start my_cluster. Also added the necessary --irl CLI flag when executing against a running production cluster to keep vagabond from destroying the nodes at run-end. :)