Operator-sdk: Operator-sdk's test-framework fails to initialize cluster resource: Namespace

[pratikjagrut]

Bug Report

What did you do?
Tried to run e2e tests on memcached sample operator.

• For curiosity, I changed the test name and made sure it will exceed the openshift's project/namespace character limit which is 63 characters
• Then ran this command operator-sdk test local ./test/e2e --image quay.io/pratikjagrut/memcached-operator:v0.0.2 --verbose

What did you expect to see?
SDK to handle exceeded namespace character limit and tests to pass.

What did you see instead? Under which circumstances?
Fails with the following error.

DEBU[0000] Debug logging is set                         
INFO[0000] Testing operator locally.                    
DEBU[0000] Running []string{"go", "test", "./test/e2e/...", "-namespacedMan", "/tmp/namespaced-manifest.yaml605350766", "-globalMan", "/tmp/global-manifest.yaml069236213", "-root", "/home/pjagrut/go/src/github.com/operator-framework/operator-sdk-samples/memcached-operator"} 
=== RUN   TestMemcached
=== RUN   TestMemcached/memcached-group
=== RUN   TestMemcached/memcached-group/Cluster-memcached-memcached-group-cluster
=== PAUSE TestMemcached/memcached-group/Cluster-memcached-memcached-group-cluster
=== CONT  TestMemcached/memcached-group/Cluster-memcached-memcached-group-cluster
--- FAIL: TestMemcached (27.18s)
    --- FAIL: TestMemcached/memcached-group (0.00s)
        --- FAIL: TestMemcached/memcached-group/Cluster-memcached-memcached-group-cluster (0.31s)
            memcached_test.go:98: failed to initialize cluster resources: Namespace "memcached-memcached-group-cluster-memcached-memcached-group-cluster-1571914599" is invalid: metadata.name: Invalid value: "memcached-memcached-group-cluster-memcached-memcached-group-cluster-1571914599": must be no more than 63 characters
FAIL
FAIL    github.com/operator-framework/operator-sdk-samples/memcached-operator/test/e2e  54.554s
Error: failed to build test binary: (failed to exec []string{"go", "test", "./test/e2e/...", "-namespacedMan", "/tmp/namespaced-manifest.yaml605350766", "-globalMan", "/tmp/global-manifest.yaml069236213", "-root", "/home/pjagrut/go/src/github.com/operator-framework/operator-sdk-samples/memcached-operator"}: exit status 1)
Usage:
  operator-sdk test local <path to tests directory> [flags]

Flags:
      --debug                         Enable debug-level logging
      --global-manifest string        Path to manifest for Global resources (e.g. CRD manifests)
      --go-test-flags string          Additional flags to pass to go test
  -h, --help                          help for local
      --image string                  Use a different operator image from the one specified in the namespaced manifest
      --kubeconfig string             Kubeconfig path
      --local-operator-flags string   The flags that the operator needs (while using --up-local). Example: "--flag1 value1 --flag2=value2"
      --molecule-test-flags string    Additional flags to pass to molecule test
      --namespace string              If non-empty, single namespace to run tests in
      --namespaced-manifest string    Path to manifest for per-test, namespaced resources (e.g. RBAC and Operator manifest)
      --no-setup                      Disable test resource creation
      --up-local                      Enable running operator locally with go run instead of as an image in the cluster

Global Flags:
      --verbose   Enable verbose logging

Environment

• operator-sdk version: v0.10.0-117-gc7f429f0", commit: "c7f429f05808406068b1a67f49b0de9e6481f545

• go version: go1.12.9 linux/amd64

• Kubernetes version information:

Client Version: version.Info{Major:"1", Minor:"10+", GitVersion:"v1.10.0+d4cacc0", GitCommit:"d4cacc0", GitTreeState:"clean", BuildDate:"2018-12-06T15:15:06Z", GoVersion:"go1.11.2", Compiler:"gc", Platform:"linux/amd64"}
Server Version: version.Info{Major:"1", Minor:"16+", GitVersion:"v1.16.0-beta.2+03b33d2", GitCommit:"03b33d2", GitTreeState:"clean", BuildDate:"2019-10-21T19:38:32Z", GoVersion:"go1.12.5", Compiler:"gc", Platform:"linux/amd64"}

• Are you writing your operator in ansible, helm, or go?
  GO

Possible Solution
Seems like SDK appends operator name, test-group name and test name to create the namespace without taking care of the length of the final string.
We should check the char length of the final string before creating the namespace.
If it exceeds then should cut it down to less than 63 chars or else it is good to go.

[camilamacedo86]

Hi @AlexNPavel,

I understand that the namespace is mandatory. Am I right? If yes, maybe the solution could not allow the command to go forward without it? Or should it work without that namespace flag? If yes, what about we just use the operator-name + randon numbers trunc (see the msg below)? WDYT?

[pratikjagrut]

@camilamacedo86 What I noticed is if we do not provide namespace SDK creates one by appending test-group name, test name, operator name and around 10 digits random number. So I don't see why --namespace should be compulsory if not using --up-local?

[AlexNPavel]

The reason the SDK creates namespaces with that form (group-test-timestamp) is to prevent namespace conflicts when running tests in parallel. If the namespace name is too long, I assume we could truncate the name, but then we might have namespace conflicts. The test framework errors if it tries to create a namespace that already exists, so maybe it would be fine to try and truncate the namespace's name before trying to create it and make the error message for the namespace conflicts a bit clearer.

@joelanford What are your thoughts on this?

[joelanford]

I think the best bet would be to change the code to always use a consistent and known length for the namespace name (e.g. osdk-e2e-<54-or-less-random-alphanum-chars>)

[camilamacedo86]

Hi @pratikjagrut,

Now, you have your answers if you still wishing to collab with a PR with the fix I think you can do as @joelanford suggest.

[pratikjagrut]

HI @camilamacedo86,
Thanx, I'll raise PR ASAP.