* Upgrade to kube-prometheus-stack 67.x
* Prometheus operator is upgraded to 0.80.0
* Prometheus is upgraded from 2.55.0 to 3.2.1
* Enable pod monitors to test them
* Run linting with kube-prometheus-stack enabled
* Validate all CI configs
* Fix pulsar-cluster-initialize / pulsar-init rendering with kustomize
- reapply #166 changes that were reverted by #544 changes
* Add validation for kustomize output in CI
* Fix Oxia config so that it includes a list of all pods in the statefulset
* Test Oxia with 3 replicas since some issues only come up with more nodes
* Make internal name not a fqdn
* Fix issue with insufficient cpu requests in CI
* Use Pulsar 4.0.0 image, bump chart version to 3.7.0
* Bump kube-prometheus-stack to 65.x.x
* Remove testing with latest and test with previous LTS version
- run kube-prometheus-stack test with previous LTS version since
the older chart version doesn't support Pulsar 4.0.0 image
* Fix passing "--values" to helm command
* Move ci runner config to a script
* Attempt to fix pulsar-manager-cluster-initialize
* Make zookeeper healthchecks compatible with Alpine's busybox nc
* Test Pulsar 3.3.0 image
* Use 127.0.0.1 instead of localhost in zookeeper healthchecks
- Alpine nc fails if "localhost" is used.
- perhaps it defaults to use IPv6?
* Disable testing with Pulsar 3.3.0 image until 3.3.1 is released
- the image needs "apk add bind-tools" since busybox nslookup isn't compatible with kubernetes
* Added support for JWT secretref and key volume mount. Added admin user auto-creation.
* Removed variables accidentally re-added and comments no longer relevant
* Enabling pulsar manager test w/ manager-admin superuser for symmetric and asymmetric jwt tests
* Added verification of communication with broker to ci test-pulsar-manager
* Fixing error on line 115 of helm.sh
* More fixes
* Adding echo of envs and tenants
* Fixing LOGIN_JSESSIONID variable name
* Add upgrade test to CI
- install latest released chart version and then upgrade to the local chart version
* Add metallb for loadbalancer support so that "--wait" will work
* Improve script logging
* produce messages with old version of pulsar and consume with new version
* Add 10 second delay
* Remove duplication
* Improve logging
* Fix quotes
* Before trying produce-consume test, verify that the Admin API is available
* Add 2 more upgrade scenarios: TLS & PSP
* Proposal: service accounts creation should be decoupled from PodSecurityPolicy.
* Rename *-rbac.yaml to *-psp.yaml and move service account to *-service-account.yaml
* Test with psp enabled
Co-authored-by: Lari Hotari <lhotari@apache.org>
* Upgrade upgrade kind, chart releaser and helm versions
* Disable podMonitory for values-broker-tls.yaml file
- was missing from #317
* Use k8s 1.18.20
* Use ubuntu-20.04 runtime
- k8s < 1.19 doesn't support cgroup v2
* Upgrade to k8s 1.19 as baseline
* Baseline to k8s 1.20
* Set ip family to ipv4
* Add more logging to kind cluster creation
* Simplify duplicate job deletion
* use verbosity flag
* Upgrade to k8s 1.24
* Replace removed tolerate-unready-endpoints annotation with publishNotReadyAddresses
(cherry picked from commit e90926053a2b01bb95529fbaddc8d2ce2cdeec63)
* Use k8s 1.21 as baseline
* Run on ubuntu-22.04
* Use Pulsar 2.10.4
* Imrpove documentation and testing for PodMonitors
* Fix missed references and a typo
### Motivation
Before upgrading to 3.0.0, we want to make sure the kube-prometheus-stack is well documented.
### Modifications
* Update tests and examples to fully disable `PodMonitors` and the installation of the kube-prometheus-stack CRDs.
### Verifying this change
The current tests will cover these changes.
### Motivation
The current tests only produce a message. This test adds a consumer for the produced message.
### Modifications
* Add new section to the test where we consume the produced message
Related to #311
### Motivation
In #299, I updated the values without also updating the test values. As a result, I unintentionally enabled the monitoring stack in the tests and broke some examples. Because we are deploying all resources to a single node. It is possible that we are resource constrained, so I am going to re-disable the monitoring stack.
### Modifications
* Update test cluster configurations to re-disable deploying the monitoring stack
* Update examples with the new configuration
### Verifying this change
- [ ] Make sure that the change passes the CI checks.
* Replace monitoring solution with kube-prometheus-stack dependency
* Enable pod monitors
* Download necessary chart dependencies for CI
* Actually run dependency update
* Enable missed podMonitor
* Disable alertmanager by default for feature parity
Related issues #294#65
Supersedes #296 and #297
### Motivation
Our helm chart is out of date. I propose we make a breaking change for the monitoring solution and start using the `kube-prometheus-stack` as a dependency. This should make upgrades easier and will let users leverage all of that chart's features.
This change will result in the removal of the StreamNative Grafana Dashboards. We'll need to figure out the right way to address that. The apache/pulsar project has grafana dashboards, but they have not been maintained. With this added dependency, we'll have the benefit of being able to use k8s `ConfigMap`s to configure grafana dashboards.
### Modifications
* Remove old prometheus and grafana configuration
* Add kube-prometheus-stack chart as a dependency
* Enable several components by default. I am not opinionated on these, but it is based on the other values in the chart.
### Verifying this change
This is a large change that will require manual validation, and may break deployments. I propose this triggers a helm chart 3.0.0 release.
* Use cert-manager to generate certs for tests
* Install Cert-Manager in test env
### Motivation
Currently, we use hard coded certificates for the tests. Instead, we can use Cert Manager to generate the certificates. The primary benefit of this change is that it ensure we're testing the cert manager integration.
### Modifications
* Remove `.ci/tls` directory since we no longer need these certs.
* Remove `scripts/pulsar/clean_tls.sh` (it wasn't used)
* Remove `scripts/pulsar/upload_tls.sh` since we are not uploading any certs
* Update the `helm.sh` test script
* Update the `.ci/clusters` configurations to generate the relevant cert manager manifests
### Verifying this change
- [ ] Make sure that the change passes the CI checks.
* Bump Apache Pulsar 2.10.1
* Do not bump .Chart.version
* Remove unnecessary jq download that was failing with Permission Denied
Co-authored-by: Michael Marshall <mmarshall@apache.org>
Relates to: https://github.com/apache/pulsar-helm-chart/issues/290
### Motivation
We should not use GitHub Actions to release the helm chart. As such, we can remove the relevant workflow code from this repo while we build the relevant process to officially release the helm chart.
The main risk with this kind of change is that we won't have a way to "release" the chart. However, it is relevant to point out that we have not had any official releases of the chart given that the PMC has not been voting on the releases. I think we need to prioritize fixing this process as a community.
### Modifications
* Remove all scripts and configuration files that enabled GitHub Actions to release the helm chart.
### Verifying this change
This is a trivial change.
* Bump version to `2.9.2`
* Because the latest Pulsar image is based on Java 11, some JVM param for printing GC information has been abandoned, change to use the new JVM param. refer to https://docs.oracle.com/en/java/javase/11/tools/java.html#GUID-BE93ABDC-999C-4CB5-A88B-1994AAAC74D5 and https://issues.redhat.com/browse/CLOUD-3040.
original param | new param
--|--
`-XX:+PrintGCDetails` | `-Xlog:gc*`
`-XX:+PrintGCApplicationStoppedTime` | `-Xlog:safepoint`
`-XX:+PrintHeapAtGC` | `-Xlog:gc+heap=trace`
`-XX:+PrintGCTimeStamps` | `-Xlog:gc::utctime`
* remove JVM param `-XX:G1LogLevel=finest`
