fix(secgroups): use standalone secgroup rules instead of in-line rules

[metral]

fixes: #106
rel: #69

[metral]

Note: this PR currently sets the ingress and egress prop of both secgroups to an empty array []. This is necessary as there seems to be a bug when removing both props all-together and pulumi not triggering a removal of the rules as expected. e.g. in the changes below, neither set of ingress or egress rules that were in-line were set for removal when attempting to remove props entirely. The only work around is to set each prop to an empty array instead, along with the standalone secgroup rules as done in this PR.

I'll be opening a separate issue to track this props removal bug in pulumi/pulumi.
Update: issue has been opened.

Changes:
 
    Type                           Name                                   Operation
~   aws:ec2:SecurityGroup          kx-eks-cluster-eksClusterSecurityGroup updated
~   aws:ec2:SecurityGroup          kx-eks-cluster-nodeSecurityGroup       updated
~   pulumi-nodejs:dynamic:Resource kx-eks-cluster-vpc-cni                 updated
 
Resources:
    ~ updated 3
    17 unchanged
 
Duration: 4s

[metral]

@lukehoban PTAL

[lukehoban]

LGTM.

This does make me really want some tests to ensure that this update can safely be performed from a stack deployed using previous versions of the library. Do you know exactly what happens during the update? Is there a very brief period of outage where the security groups are removed before they get added? If so, may want to add a note about the upgrade to the CHANGELOG?

[metral]

@lukehoban agree on the fact that we should enable testing prev versions against the new commits.

Do you know exactly what happens during the update?

I'd imagine that there can be a brief period of unavailability given that the secgroup rules are against the worker instances themselves. That said, this patch update does happen in a matter of seconds (< 10 sec and closer to 5sec) so it isn't a prolonged period.

I'll make sure to mention this ^ as a part of the changelog.

For the sake of more data on this change:

• here's the output for the update: https://gist.github.com/metral/c7fe7a78a78469688b57ec9b8572cde7
• here's the rich diff output for the update:
  https://gist.github.com/metral/0a702d2d262b72f21046ee448456f232