mirror of
https://github.com/fluxcd/flagger.git
synced 2026-02-23 22:34:11 +00:00
7242fa7d5c
In Linkerd 2.13 the Prometheus instance in the `linkerd-viz` namespace is now locked behind an [_AuthorizationPolicy_](https://github.com/linkerd/linkerd2/blob/stable-2.13.1/viz/charts/linkerd-viz/templates/prometheus-policy.yaml) that only allows access to the `metrics-api` _ServiceAccount_. This adds an extra _AuthorizationPolicy_ to authorize the `flagger` _ServiceAccount_. It's created by default when using Kustomize, but needs to be opted-in when using Helm via the new `linkerdAuthPolicy.create` value. This also implies that the Flagger workload has to be injected by the Linkerd proxy, and that can't happen in the same `linkerd` namespace where the control plane lives, so we're moving Flagger into the new injected `flagger-system` namespace. The `namespace` field in `kustomization.yml` was resetting the namespace for the new _AuthorizationPolicy_ resource, so that gets restored back to `linkerd-viz` using a `patchesJson6902` entry. A better way to do this would have been to use the `unsetOnly` field in a _NamespaceTransformer_ (see kubernetes-sigs/kustomize#4708) but for the life of me I couldn't make that work... Signed-off-by: Alejandro Pedraza <alejandro@buoyant.io>
19 lines
495 B
Bash
Executable File
19 lines
495 B
Bash
Executable File
#!/usr/bin/env bash
|
|
|
|
set -o errexit
|
|
|
|
REPO_ROOT=$(git rev-parse --show-toplevel)
|
|
DIR="$(cd "$(dirname "$0")" && pwd)"
|
|
|
|
"$DIR"/install.sh
|
|
|
|
"$REPO_ROOT"/test/workloads/init.sh
|
|
# Delete Daemonset as it eats up precious CPU requests and we don't need it anyway.
|
|
kubectl -n test delete ds podinfo-ds
|
|
"$DIR"/test-canary.sh
|
|
|
|
"$REPO_ROOT"/test/workloads/init.sh
|
|
# Delete Daemonset as it eats up precious CPU requests and we don't need it anyway.
|
|
kubectl -n test delete ds podinfo-ds
|
|
"$DIR"/test-steps.sh
|