Some providers support adding tags to test. We should allow for additional information which will help with filtering.

Add a mapping so we can create checks with StatusCake.

There is a provider interface, which this mapper should follow. We can then register this provider within an init function and importing it in our command.

Currently the LabelSelector doesn't support matchExpression. We should implement this using the metav1.LabelSelectorAsSelector function.

Implementing this will be easier when we use cache informers.

Set up a docker hub image and hook it up to Travis

The current docs are outdated, we need to explain the Monitor/MonitorTemplate differences.

It should also contain a correct deployment with the operator, RBAC and CRD objects (with validation rules)

We're missing a lot of validation rules. We need to set this up.

I'd like to see both client side (CRD Spec) and server side (operator) validation.

We currently have a very basic test. We should add a way more specific test to test changing behaviour.

Instead of querying the API constantly, we should use a cache informer to give us the information we want.

This will help reduce stress on master nodes for clusters where this is deployed.

When an Ingress has multiple hosts, there will be multiple monitors created. Let's figure out a way that we can simplify this and add an option for one host per Ingress object.

Removing a label from the ingress doesn't seem to trigger removal of the IngressMonitor, resulting in the IngressMonitor still being present and the StatusCheck still being available on the Provider.

When an Ingress changes, we should validate if it is still selected by a Monitor. If not, we need to find all IngressMonitors which are owned by the Ingress and delete them.

A resource can't have multiple ControllerReferences where Controller = true.

This configuration however is responsible for making sure an item is Garbage Collected. Since we want to make sure that when either a Monitor or Ingress is deleted, the underlying IngressMonitor gets deleted as well, we need to set up a mechanism to do this.

We can do this by using the OnDelete listener and add some GC capabilities in there.

We should have a metrics endpoint which can be used by Prometheus to scrape information.

• number of providers
• number of templates
• number of monitors
• number of ingress monitors
• number of failed syncs

The operator should log out more information about successful transactions.

• successfully created a monitor
• successfully deleted a monitor
• successfully updated a monitor

Instead of manually creating the CRD manifests, we should pull this out of the apis folder and use that to release the actual manifests. Add it with the make generated command.

We should set up a healthz endpoint to indicate that the service is running fine. This should be done after all mandatory checks so we can retry if needed.

The StatusCake client isn't providing that much information about errors. We should swap to our own integration and do status code checks etc. to see if we've got duplicated requests or if we're trying to update a resource which doesn't exist.

It is also lacking support for SSL checks, which we need to enable.

In some multi-tenant or multi-environment clusters, it would be useful to have namespace scoped providers and monitortemplates, paired with an operator that runs within that namespace only.

We should make a checksum of the IngressRule Host and append that to the name of the generated IngressMonitor when adding a Monitor.

This way we ensure uniqueness across multiple rules.