operator: Add e2e tests for DSA, IAA

[ozhuraki]

No description provided.

[codecov-commenter]

Codecov Report

Merging #1196 (0e49e48) into main (cb3b2a7) will not change coverage.
The diff coverage is n/a.

❗ Current head 0e49e48 differs from pull request most recent head ef7954c. Consider uploading reports for the commit ef7954c to get more accurate results

@@           Coverage Diff           @@
##             main    #1196   +/-   ##
=======================================
  Coverage   50.78%   50.78%           
=======================================
  Files          41       41           
  Lines        4608     4608           
=======================================
  Hits         2340     2340           
  Misses       2133     2133           
  Partials      135      135           

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[mythi]

Internal error occurred: failed calling webhook "webhook.cert-manager.io": failed to call webhook: Post "https://cert-manager-webhook.cert-manager.svc:443/mutate?timeout=10s": context deadline exceeded

@ozhuraki I've seen this error for the past ~2 rounds of your pushesh. Interestingly, e2e-iaa is OK... something with the operator manager deployment seems to trigger these. Just making wild guesses but long time ago setting GOMAXPROCS=1 env variable to manager helped with some cpu throttling problems.

[ozhuraki]

Changes:

• Added a check for NFD readiness
• Removed CPU limits for the operator under e2e
• Increased DSA, IAA webhook timeouts
• Added GOMAXPROCS to the operator e2e deployment
• Dropped BeforeEeach, AfterEeach in SGX (SGX's AfterEach deleted NFD).
  After the NFD was moved to BeforeSuite, this started to interfere,
  yet to occasionally producing an OK overal pass

[mythi]

Changes:

* Added a check for NFD readiness

The NFD rework can go to a PR of its own. The same for the initImage additions to DSA/IAA resource YAMLs.

* Removed CPU limits for the operator under e2e

* Increased DSA, IAA webhook timeouts

* Added GOMAXPROCS to the operator e2e deployment

Do we understand the problem and why these (and the unconditional 2 minutes sleep) are needed?

* Dropped BeforeEeach, AfterEeach in SGX (SGX's AfterEach deleted NFD).
  After the NFD was moved to BeforeSuite, this started to interfere,
  yet to occasionally producing an OK overal pass

Yes this is the right thing to do. I wonder if we have similar race conditions elsewhere...

[mythi]

a couple of comments before the rebase.

[ozhuraki]

Changes:

• Dropped a timeout after the operator deployment
• Added a check for the webhook availability
  Apparently a simple curl check (--connect-timeout ...) isn't enough.
  Despite the check succeeding, webhook occasionally timeouts at CR creation
  (even with e2e customizations: increased timeout, no CPU limit).
• Dropped operator e2e customizations

[ozhuraki]

Changes:

• Changed the order of applied customizations in the default operator deployment
  preventing a clean undeployment.
  Apparently this is a long standing bug leading to sporadic errors
  on operator undeployment (even in the simplest scenario: manual operator deploy/undeploy)
• Added the deletion of deployed operator resources
• Increased the terminationGracePeriodSeconds.
  The default also results in errors on undeployment
• Dropped the namespace from the deletion kubectl (deployments/operator/default)
  This also results in error under k8s/e2e

[mythi]

• Changed the order of applied customizations in the default operator deployment
  preventing a clean undeployment.
  Apparently this is a long standing bug leading to sporadic errors
  on operator undeployment (even in the simplest scenario: manual operator deploy/undeploy)

kustomize orders its output. What is the error you are seeing? I tried kubectl kustomize with and without this change and the output yaml is the same.

[ozhuraki]

Changes:

• Dropped the changes in the order of the operator customizations

[ozhuraki]

Changes:

• Dropped a namespace at CR creation / deletion

[ozhuraki]

Changes:

• Dropped a TLS check
• Added AddReadyzCheck() in the operator
  Had a false negative with this earlier

[mythi]

it's looking good. a couple of questions still

[ozhuraki]

Changes:

• Dropped a namespace when deploying the operator
• Removed the deletion of the namespace
• Corrected a previous mistake with AddReadyzCheck()

[mythi]

@ozhuraki thanks! we're finally past the simics woes and the end result looks good