jboyd01

You can validate Service Catalog using a couple of different approaches.  The Service Catalog E2E is one (see the instructions 
for syncing Service Catalog to OpenShift).  Additionally, you can utilize the OpenShift web console or the CLI and add 
applications exposed by the Template Service Broker or Ansible Service Broker.  Note that some Templates create applications 
directly and some use the Service Catalog.  Python & HTTPD don't use Service Catalog, most templates with "persistent" in the 
name do.  For those that don't, there is no ServiceInstance, Class or binding created.  Once you add applications to your 
project, you should be able to see them via the CLI (oc get serviceinstances --all-namespaces)

In the OpenShift web console (https://127.0.0.1:8443) login as Developer (any password works) and select MyProject (or any 
other project and then "Add to Project", "Browse Catalog".  I believe the Ansible Service Broker offerings have (APB) in the 
name. 

jboyd01

We want to expose Metrics from Catalog via prometheus to enable monitoring and track key metrics and provide the ability 
to alert on specific conditions.

Prometheus provides a client api that enables you to register a HTTP handler (ie /metrics) that automatically exposes 
Prometheus metrics objects.  Many core components within Kubernetes already do this including the CAdvisor, Kubelet, 
Scheduler, Proxy,  and many more.  The Prometheus server can be easily configured with scrap configurations that will 
poll the /metrics endpoints and provide a centralized UI for discovery & analysis.  Advanced analytic and graphing tools
such as Grafana can consume Prometheus data are often used to augment monitoring.

Kubernetes API Servers are exposing Prometheus metrics out of the box, Prometheus must be 

jboyd01

my configuration specified:
openshift_version: latest
openshift_image_tag: latest
openshift_pkg_version: "3.6.0"

I checked out release-3.6.0 for the openshift-ansible repository prior to running.

I ran $ ansible-playbook /home/jaboyd/go/src/github.com/openshift-ansible/playbooks/byo/config.yml \
   -i hosts  \
   -e @svc-cat-install

jboyd01

Purpose
Allow brokers to register callbacks (webhooks) that can be used for validation prior to the platform attempting to Create, Update or Delete (CUD) a Service Instance or Binding.

There will likely be a lot of discussion on the actual implementation details, initially this proposal will just focus on surfacing the issue and proposing the use of pre-action validation so Brokers have an opportunity to indicate to the platform that an action will or will not be accepted for processing.  Once the SIG has discussed and given general agreement, we'll drill into a detailed design.

This feature allows a broker to register webhooks for precheck validation for Instances and Bindings.  That is, if indicated by the broker, the Platform will invoke a validating webhook just prior to invoking the actual call to create, update or delete an Instance or Binding.  The webhook will be invoked with the same parameters and payload as the actual create/update/delete operation, but this operation is a dry-run for Broker vali

jboyd01

Spec:
  Cluster Service Class External Name:  example-starter-pack-service
  Cluster Service Class Ref:
    Name:                              4f6e6cf6-ffdd-425f-a2c7-3c9258ad246a
  Cluster Service Plan External Name:  default
  Cluster Service Plan Ref:
    Name:       86064792-7ea2-467b-af93-ac9694d96d5b
  External ID:  20da0666-a570-11e8-97f3-0242ac110003

jboyd01

I've got a BZ issue that indicates events are sometimes missing on service
instances.  I've been able to replicate it.  Generally you see all events, but
sometimes,  it seems like the most recent events are not displayed.  For
instance, in my controller log I pull out all occurrences of "event.go" which
shows the events that were set on (in this case) the instance.  The instance was
in an error condition for a bit (set to an invalid plan), but then the last
action I did updated the plan to a valid plan:


21:41:58.603293       1 event.go:221] Event(v1.ObjectReference{Kind:"ServiceInstance", Namespace:"default", Name:"myservice2",...... ResourceVersion:"126", FieldPath:""}): type: 'Warning' reason: 'ReferencesNonexistentServicePlan' References a non-existent ClusterServicePlan {ClusterServicePlanExternalName:"defaultXXX"} on ClusterServiceClass 4f6e6cf6-ffdd-425f-a2c7-3c9258ad246a {ClusterServiceClassExternalName:"example-starter-pack-service"} or there is more than one (found: 0)

jboyd01

cat <<'EOF' | oc create -f -
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRole
metadata:
  name: add-servicebindingfinalizers
rules:
- apiGroups:
  - servicecatalog.k8s.io
  resources:
  - servicebindings/finalizers

jboyd01

#!/bin/bash

for p in $(oc get pods -n kube-service-catalog -o name -l app=apiserver); do
    pod=$(sed -e 's/pod\///g'<<<$p)
    oc logs $pod -c apiserver -n kube-service-catalog > /tmp/artifacts/$pod.log
done
for p in $(oc get pods -n kube-service-catalog -o name -l app=controller-manager); do
    pod=$(sed -e 's/pod\///g'<<<$p)
    oc logs $pod -n kube-service-catalog > /tmp/artifacts/$pod.log
done

jboyd01

apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRole
metadata:
  name: system:service-catalog:aggregate-to-admin
  labels:
    rbac.authorization.k8s.io/aggregate-to-admin: "true"
rules:
- apiGroups:
  - "servicecatalog.k8s.io"
  attributeRestrictions: null

jboyd01

As of Feb 24 Service Catalog is now installed by two new Cluster Operators. Initially Service Catalog is not enabled/installed. To enable it, the cluster admin must create two custom resources as follows:

cat <<EOF | oc create -f -
apiVersion: operator.openshift.io/v1
kind: ServiceCatalogAPIServer
metadata:
  name: cluster
spec:
 logLevel: "Normal"