As suggested in #339, this PR updates Flintrock so that it tags launched nodes as part of the launch itself, rather than in a separate call post-launch. This means that we can always rely on tags to find cluster nodes, even when a launch operation fails midway.
This PR also:
Adjusts when and how long we wait for AWS to process launch requests before proceeding with other operations.
Tweaks the message Flintrock prints when launching nodes.
Makes some formatting changes. (At this point I am inclined to reformat the entire repo with black or something similar.)
I tested this PR by launching and destroying clusters, and interrupting cluster launches at various points to test Flintrock's ability to find the half-baked nodes.
As suggested in #339, this PR updates Flintrock so that it tags launched nodes as part of the launch itself, rather than in a separate call post-launch. This means that we can always rely on tags to find cluster nodes, even when a launch operation fails midway.
This PR also:
black
or something similar.)I tested this PR by launching and destroying clusters, and interrupting cluster launches at various points to test Flintrock's ability to find the half-baked nodes.
Fixes #183. Fixes #339. Related to #179.