allow set tolerate failed pods for each step

Signed-off-by: mingzhou.swx <[email protected]>

api/v1alpha1/batchrelease_plan_types.go

@@ -52,6 +52,16 @@ type ReleaseBatch struct {
 	// it can be an absolute number (ex: 5) or a percentage of workload replicas.
 	// batches[i].canaryReplicas should less than or equal to batches[j].canaryReplicas if i < j.
 	CanaryReplicas intstr.IntOrString `json:"canaryReplicas"`
+	// MaxUnavailable indicates how many failed replicas can be tolerated in current batch.
+	// For example, assume that 10 pods are upgraded in this batch, and MaxUnavailable=20%,
+	// this batch can enter ready state if at latest 8 upgraded pods are ready.
+	// Specifically, Rollout will also calculate a global toleration according to all the upgraded
+	// pods and the MaxUnavailable in workload. The batch should satisfy both of them before enter
+	// the next batch.
+	// MaxUnavailable will take effect only when RolloutID is set.
+	// Default to nil.
+	// +optional
+	MaxUnavailable *intstr.IntOrString `json:"maxUnavailable,omitempty"`
 }
 
 // BatchReleaseStatus defines the observed state of a release plan

api/v1alpha1/rollout_types.go

@@ -101,8 +101,7 @@ type CanaryStrategy struct {
 
 // CanaryStep defines a step of a canary workload.
 type CanaryStep struct {
-	// SetWeight sets what percentage of the canary pods should receive
-
+	// Weight indicate how many percentage of traffic the canary pods should receive
 	// +optional
 	Weight *int32 `json:"weight,omitempty"`
 	// Replicas is the number of expected canary pods in this batch
@@ -111,6 +110,16 @@ type CanaryStep struct {
 	// Pause defines a pause stage for a rollout, manual or auto
 	// +optional
 	Pause RolloutPause `json:"pause,omitempty"`
+	// MaxUnavailable indicates how many failed replicas can be tolerated in current batch.
+	// For example, assume that 10 pods are upgraded in this batch, and MaxUnavailable=20%,
+	// this batch can enter ready state if at latest 8 upgraded pods are ready.
+	// Specifically, Rollout will also calculate a global toleration according to all the upgraded
+	// pods and the MaxUnavailable in workload. The batch should satisfy both of them before enter
+	// the next batch.
+	// MaxUnavailable will take effect only when RolloutID is set.
+	// Default to nil.
+	// +optional
+	MaxUnavailable *intstr.IntOrString `json:"maxUnavailable,omitempty"`
 	// MetricsAnalysis *RolloutAnalysis `json:"metricsAnalysis,omitempty"`
 }
 

config/crd/bases/rollouts.kruise.io_batchreleases.yaml

@@ -88,6 +88,21 @@ spec:
                             should less than or equal to batches[j].canaryReplicas
                             if i < j.'
                           x-kubernetes-int-or-string: true
+                        maxUnavailable:
+                          anyOf:
+                          - type: integer
+                          - type: string
+                          description: MaxUnavailable indicates how many failed replicas
+                            can be tolerated in current batch. For example, assume
+                            that 10 pods are upgraded in this batch, and MaxUnavailable=20%,
+                            this batch can enter ready state if at latest 8 upgraded
+                            pods are ready. Specifically, Rollout will also calculate
+                            a global toleration according to all the upgraded pods
+                            and the MaxUnavailable in workload. The batch should satisfy
+                            both of them before enter the next batch. MaxUnavailable
+                            will take effect only when RolloutID is set. Default to
+                            nil.
+                          x-kubernetes-int-or-string: true
                       required:
                       - canaryReplicas
                       type: object

config/crd/bases/rollouts.kruise.io_rollouts.yaml

@@ -98,6 +98,21 @@ spec:
                         items:
                           description: CanaryStep defines a step of a canary workload.
                           properties:
+                            maxUnavailable:
+                              anyOf:
+                              - type: integer
+                              - type: string
+                              description: MaxUnavailable indicates how many failed
+                                replicas can be tolerated in current batch. For example,
+                                assume that 10 pods are upgraded in this batch, and
+                                MaxUnavailable=20%, this batch can enter ready state
+                                if at latest 8 upgraded pods are ready. Specifically,
+                                Rollout will also calculate a global toleration according
+                                to all the upgraded pods and the MaxUnavailable in
+                                workload. The batch should satisfy both of them before
+                                enter the next batch. MaxUnavailable will take effect
+                                only when RolloutID is set. Default to nil.
+                              x-kubernetes-int-or-string: true
                             pause:
                               description: Pause defines a pause stage for a rollout,
                                 manual or auto
@@ -117,6 +132,8 @@ spec:
                                 5) or a percentage of total pods.'
                               x-kubernetes-int-or-string: true
                             weight:
+                              description: Weight indicate how many percentage of
+                                traffic the canary pods should receive
                               format: int32
                               type: integer
                           type: object