* After repo move to helm/helm, update steps for Owning and Maintaining A Chart
* Change repo URL from kubernetes/helm to helm/helm everywhere above chart directories (those may be separate PRs)
* Fix earlier missing apostrophe and incorrect case
* Add missing k8s/ > helm/ string changes
* [CI] Refactor repo sync
With the upcoming refactoring of the chart testing, Shellcheck is added
to CircleCI for shell scripts. In order to prepare for this, several
adjustments are necessary so it passes. On top of that, the script is
refactored as follows:
* Reorganize code into several functions for better readability and to
eliminate code redundancies
* Use curl instead of wget for intalling Helm because it's already
available in the Docker image
* Treat index.yaml separately and copy it to the bucket after syncing
charts in order to avoid chart fetch errors
* Fix repo bucket urls
* Only log errors but don't exit with non-zero exit code
* Enhancing lint.sh with additional checks
* reverting post local testing changes
* addressing reviews.
removing yq installetion
chaning yq with yaml for Chart.yaml parsing.
modifing Erro message for not github account check.
also modifing lint.sh so it can be used locally
1. BUILD_NUMBER is deprecated and needed to be replaced for prow jobs
2. PULL_NUMBER is not available when batch jobs, including merges,
are executed. Need a fallback to avoid errors.
* Fixing orphaned helm deploys left by CI
The CI system has cases where helm delete and namespace cleanup
are skipped if there is an error getting logs on a pod. This causes
the pod logs to always return true and avoid a premature exit before
cleanup is run
* Fixing issue where pods not yet running to not have delay before rechecking
When pods are not yet in a running state there is no delay before
rechecking. This is causing timeouts to happen waiting for pods.
This fixes flaky readiness checks by double-checking readiness. It could
happen that the script deemed all pods ready when at the moment the
check is executed only ready pods are reported but not all replicas are
running yet.
This change does 3 things:
1. Gets the logs from the test pods
2. Prints the output of helm status
3. Gets the logs from the pods running for a release
When the array splitting happen there are cases where the patch
(maj.min.patch) is 0. Using ++ to increment that 0 causes a
return code of 1. Using +=1 instead increments the value but has
a return code of 0.
The bot that merges PRs re-runs the e2e job before merging a pull
request. By also having it here (circleci provides fast response
to PR authors) we make sure the version has been incremented in
the chart prior to it being merged.
