Add time-limited retries to InstallPlan execution.

cmd/catalog/main.go

@@ -69,6 +69,8 @@ var (
 
  profiling = flag.Bool(
  "profiling", false, "serve profiling data (on port 8080)")
+
+ installPlanTimeout = flag.Duration("install-plan-retry-timeout", 1*time.Minute, "time since first attempt at which plan execution errors are considered fatal")
 )
 
 func init() {
@@ -173,7 +175,7 @@ func main() {
  }
 
  // Create a new instance of the operator.
- op, err := catalog.NewOperator(ctx, *kubeConfigPath, utilclock.RealClock{}, logger, *wakeupInterval, *configmapServerImage, *utilImage, *catalogNamespace, k8sscheme.Scheme)
+ op, err := catalog.NewOperator(ctx, *kubeConfigPath, utilclock.RealClock{}, logger, *wakeupInterval, *configmapServerImage, *utilImage, *catalogNamespace, k8sscheme.Scheme, *installPlanTimeout)
  if err != nil {
  log.Panicf("error configuring operator: %s", err.Error())
  }

deploy/chart/crds/0000_50_olm_00-clusterserviceversions.crd.yaml

@@ -4527,6 +4527,19 @@ spec:
  type: string
  url:
  type: string
+ relatedImages:
+ description: List any related images, or other container images that your Operator might require to perform their functions. This list should also include operand images as well. All image references should be specified by digest (SHA) and not by tag. This field is only used during catalog creation and plays no part in cluster runtime.
+ type: array
+ items:
+ type: object
+ required:
+ - image
+ - name
+ properties:
+ image:
+ type: string
+ name:
+ type: string
  replaces:
  description: The name of a CSV this one replaces. Should match the `metadata.Name` field of the old CSV.
  type: string
@@ -4560,6 +4573,11 @@ spec:
  type: object
  additionalProperties:
  type: string
+ skips:
+ description: The name(s) of one or more CSV(s) that should be skipped in the upgrade graph. Should match the `metadata.Name` field of the CSV that should be skipped. This field is only used during catalog creation and plays no part in cluster runtime.
+ type: array
+ items:
+ type: string
  version:
  description: OperatorVersion is a wrapper around semver.Version which supports correct marshaling to YAML and JSON.
  type: string

deploy/chart/crds/0000_50_olm_00-installplans.crd.yaml

@@ -206,6 +206,9 @@ spec:
  type:
  description: InstallPlanConditionType describes the state of an InstallPlan at a certain point as a whole.
  type: string
+ message:
+ description: Message is a human-readable message containing detailed information that may be important to understanding why the plan has its current status.
+ type: string
  phase:
  description: InstallPlanPhase is the current status of a InstallPlan as a whole.
  type: string
@@ -249,6 +252,10 @@ spec:
  status:
  description: StepStatus is the current status of a particular resource an in InstallPlan
  type: string
+ startTime:
+ description: StartTime is the time when the controller began applying the resources listed in the plan to the cluster.
+ type: string
+ format: date-time
  served: true
  storage: true
  subresources:

deploy/ocp/values.yaml

@@ -18,17 +18,17 @@ olm:
  kubernetes.io/os: linux
  node-role.kubernetes.io/master: ""
  tolerations:
- - key: node-role.kubernetes.io/master
- operator: Exists
- effect: NoSchedule
- - key: node.kubernetes.io/unreachable
- operator: Exists
- effect: NoExecute
- tolerationSeconds: 120
- - key: node.kubernetes.io/not-ready
- operator: Exists
- effect: NoExecute
- tolerationSeconds: 120
+  - key: node-role.kubernetes.io/master
+  operator: Exists
+  effect: NoSchedule
+  - key: node.kubernetes.io/unreachable
+  operator: Exists
+  effect: NoExecute
+  tolerationSeconds: 120
+  - key: node.kubernetes.io/not-ready
+  operator: Exists
+  effect: NoExecute
+  tolerationSeconds: 120
  tlsCertPath: /var/run/secrets/serving-cert/tls.crt
  tlsKeyPath: /var/run/secrets/serving-cert/tls.key
  resources:
@@ -46,17 +46,17 @@ catalog:
  kubernetes.io/os: linux
  node-role.kubernetes.io/master: ""
  tolerations:
- - key: node-role.kubernetes.io/master
- operator: Exists
- effect: NoSchedule
- - key: node.kubernetes.io/unreachable
- operator: Exists
- effect: NoExecute
- tolerationSeconds: 120
- - key: node.kubernetes.io/not-ready
- operator: Exists
- effect: NoExecute
- tolerationSeconds: 120
+  - key: node-role.kubernetes.io/master
+  operator: Exists
+  effect: NoSchedule
+  - key: node.kubernetes.io/unreachable
+  operator: Exists
+  effect: NoExecute
+  tolerationSeconds: 120
+  - key: node.kubernetes.io/not-ready
+  operator: Exists
+  effect: NoExecute
+  tolerationSeconds: 120
  resources:
  requests:
  cpu: 10m
@@ -72,17 +72,17 @@ package:
  kubernetes.io/os: linux
  node-role.kubernetes.io/master: ""
  tolerations:
- - key: node-role.kubernetes.io/master
- operator: Exists
- effect: NoSchedule
- - key: node.kubernetes.io/unreachable
- operator: Exists
- effect: NoExecute
- tolerationSeconds: 120
- - key: node.kubernetes.io/not-ready
- operator: Exists
- effect: NoExecute
- tolerationSeconds: 120
+  - key: node-role.kubernetes.io/master
+  operator: Exists
+  effect: NoSchedule
+  - key: node.kubernetes.io/unreachable
+  operator: Exists
+  effect: NoExecute
+  tolerationSeconds: 120
+  - key: node.kubernetes.io/not-ready
+  operator: Exists
+  effect: NoExecute
+  tolerationSeconds: 120
  resources:
  requests:
  cpu: 10m