Add webhook functionality into FleetAutoscaler

Add webhook policy type into FleetAutoscaler declaration. Provided an example of a webhook pod which receives FleetAutoscaleReview with a Fleet status and based on that information calculates target replica set. This process is performed on FleetAustoscaler Sync every 30 seconds. Extends Buffer policy functionality with Webhook policy which is proposed in googleforgames#334 comments.
aLekSer · Dec 26, 2018 · 72d832e · 72d832e
1 parent 3ebf6c1
commit 72d832e
Show file tree

Hide file tree

Showing 60 changed files with 4,225 additions and 10 deletions.
diff --git a/Gopkg.lock b/Gopkg.lock
diff --git a/README.md b/README.md
@@ -45,6 +45,7 @@ Documentation and usage guides on how to develop and host dedicated game servers
  - [Create a Game Server](./docs/create_gameserver.md)
  - [Create a Game Server Fleet](./docs/create_fleet.md)
  - [Create a Fleet Autoscaler](./docs/create_fleetautoscaler.md)
+ - [Create a Webhook for FleetAutoscaler](./docs/create_webhook_fleetautoscaler.md)
  - [Edit Your First Game Server (Go)](./docs/edit_first_game_server.md)
 
 ### Guides

diff --git a/docs/create_webhook_fleetautoscaler.md b/docs/create_webhook_fleetautoscaler.md
@@ -0,0 +1,231 @@
+# Quickstart Create a Fleet Autoscaler with Webhook Policy
+
+This guide covers how you can create webhook fleet autoscaler policy.
+The main difference from the Buffer policy is that the logic on how many target replicas you need is delegated to a separate pod.
+This type of Autoscaler would send an HTTP request to the webhook endpoint every sync period (which is currently 30s) with a JSON body, and scale the target fleet based on the data that is returned.
+
+## Prerequisites
+
+It is assumed that you have read the instructions to [Create a Game Server Fleet](./create_fleet.md)
+and you have a running fleet of game servers or you could run command from Step #1.
+
+## Objectives
+
+- Deploy the Webhook Pod and service for autoscaling
+- Create a Fleet Autoscaler with Webhook policy type in Kubernetes using Agones custom resource
+- Watch the Fleet scale up when allocating GameServers
+- Watch the Fleet scale down after GameServer shutdown
+
+### 1. Deploy the fleet
+
+Run a fleet in a cluster:
+```
+kubectl apply -f https://raw.githubusercontent.com/GoogleCloudPlatform/agones/master/examples/simple-udp/fleet.yaml
+```
+
+### 2. Deploy a Webhook service for autoscaling
+
+We need to create a pod which will handle HTTP requests with json payload `FleetAutoscaleReview` and return back it with `FleetAutoscaleResponse` populated.
+
+```
+kubectl apply -f https://raw.githubusercontent.com/GoogleCloudPlatform/agones/master/examples/autoscaler-webhook/autoscaler-service.yaml
+```
+
+To check that it is running and liveness probe is fine:
+```
+kubectl describe pod autoscaler-webhook
+```
+
+```
+Name:           autoscaler-webhook-86944884c4-sdtqh
+Namespace:      default
+Node:           gke-test-cluster-default-1c5dec79-h0tq/10.138.0.2
+...
+Status:         Running
+```
+
+### 3. Create a Fleet Autoscaler
+
+Let's create a Fleet Autoscaler using the following command :
+
+```
+kubectl apply -f https://raw.githubusercontent.com/GoogleCloudPlatform/agones/master/examples/webhookfleetautoscaler.yaml
+```
+
+You should see a successful ouput similar to this :
+
+```
+fleetautoscaler.stable.agones.sev "webhook-fleet-autoscaler" created
+```
+
+This has created a FleetAutoscaler record inside Kubernetes.
+It has the link to Webhook service we deployed above.
+
+### 4. See the fleet and autoscaler status.
+
+In order to track the list of gameservers which run in your fleet you can run this command in a separate terminal tab:
+
+```
+watch "kubectl get gs -o=custom-columns=NAME:.metadata.name,STATUS:.status.state,IP:.status.address,PORT:.status.ports[0].port -n default"
+```
+
+In order to get autoscaler status use the following command:
+
+```
+kubectl describe fleetautoscaler webhook-fleet-autoscaler
+```
+
+It should look something like this:
+
+```
+Name:         webhook-fleet-autoscaler
+Namespace:    default
+Labels:       <none>
+Annotations:  kubectl.kubernetes.io/last-applied-configuration={"apiVersion":"stable.agones.dev/v1alpha1","kind":"FleetAutoscaler","metadata":{"annotations":{},"name":"webhook-fleet-autoscaler","namespace":"default...
+API Version:  stable.agones.dev/v1alpha1
+Kind:         FleetAutoscaler
+etadata:
+  Cluster Name:
+  Creation Timestamp:  2018-12-22T12:52:23Z
+  Generation:          1
+  Resource Version:    2274579
+  Self Link:           /apis/stable.agones.dev/v1alpha1/namespaces/default/fleetautoscalers/webhook-fleet-autoscaler
+  UID:                 6d03eae4-05e8-11e9-84c2-42010a8a01c9
+Spec:
+  Fleet Name:  simple-udp
+  Policy:
+    Type:  Webhook
+    Webhook:
+      Service:
+        Name:       autoscaler-webhook-service
+        Namespace:  default
+        Path:       scale
+      URL:
+Status:
+  Able To Scale:     true
+  Current Replicas:  2
+  Desired Replicas:  2
+  Last Scale Time:   <nil>
+  Scaling Limited:   false
+Events:              <none>
+```
+
+You can see the status (able to scale, not limited), the last time the fleet was scaled (nil for never), current and desired fleet size.
+
+The autoscaler make a query to a webhoook service deployed on step 1 and on response changing the target Replica size, and the fleet creates/deletes game server instances
+to achieve that number. The convergence is achieved in time, which is usually measured in seconds.
+
+### 5. Allocate Game Servers from the Fleet to trigger scale up
+
+If you're interested in more details for game server allocation, you should consult the [Create a Game Server Fleet](./create_fleet.md) page.
+Here we only interested in triggering allocations to see the autoscaler in action.
+
+```
+kubectl create -f https://raw.githubusercontent.com/GoogleCloudPlatform/agones/master/examples/simple-udp/fleetallocation.yaml -o yaml
+```
+
+You should get in return the allocated game server details, which should end with something like:
+```
+    status:
+      address: 35.247.13.175
+      nodeName: gke-test-cluster-default-1c5dec79-qrqv
+      ports:
+      - name: default
+        port: 7047
+      state: Allocated
+```
+
+Note the address and port, you might need them later to connect to the server.
+
+Run the above kubectl command one more time so that we have both servers allocated.
+
+### 6. Check new Autoscaler and Fleet status
+
+Now let's wait a few seconds to allow the autoscaler to detect the change in the fleet and check again its status
+
+```
+kubectl describe fleetautoscaler webhook-fleet-autoscaler
+```
+
+The last part should look similar to this:
+
+```
+Spec:
+  Fleet Name:  simple-udp
+  Policy:
+    Type:  Webhook
+    Webhook:
+      Service:
+        Name:       autoscaler-webhook-service
+        Namespace:  default
+        Path:       scale
+      URL:
+Status:
+  Able To Scale:     true
+  Current Replicas:  4
+  Desired Replicas:  4
+  Last Scale Time:   2018-12-22T12:53:47Z
+  Scaling Limited:   false
+Events:
+  Type    Reason            Age   From                        Message
+  ----    ------            ----  ----                        -------
+  Normal  AutoScalingFleet  35s   fleetautoscaler-controller  Scaling fleet simple-udp from 2 to 4
+```
+
+You can see that the fleet size has increased in particular case doubled to 4 gameservers, the autoscaler having compensated for the two allocated instances.
+Last Scale Time has been updated and a scaling event has been logged.
+
+Double-check the actual number of game server instances and status by running:
+
+```
+kubectl get gs -o=custom-columns=NAME:.metadata.name,STATUS:.status.state,IP:.status.address,PORT:.status.ports
+```
+
+This will get you a list of all the current `GameSevers` and their `Status > State`.
+
+```
+NAME                     STATUS      IP              PORT
+simple-udp-dmkp4-8pkk2   Ready       35.247.13.175   [map[name:default port:7386]]
+simple-udp-dmkp4-b7x87   Allocated   35.247.13.175   [map[name:default port:7219]]
+simple-udp-dmkp4-r4qtt   Allocated   35.247.13.175   [map[name:default port:7220]]
+simple-udp-dmkp4-rsr6n   Ready       35.247.13.175   [map[name:default port:7297]]
+```
+
+### 7. Check down scaling using Webhook Autoscaler policy
+
+If the fraction of allocated replicas in whole Replicas count would be less that threshold (0.3) than fleet would scale down by scaleFactor in our example 2.
+Note that example webhook server have a limitation that it would not decrease fleet replica count under `minReplicasCount` = 2.
+We need to run EXIT command on one gameserver (Use IP address and port of the allocated gameserver from the previous step) in order to decrease the number of allocated gameservers in a fleet (<0.3).
+```
+nc -u 35.247.13.175 7220
+EXIT
+```
+
+Server would be in shutdown state.
+Wait about 30 seconds.
+Then you should see in the output of next command scaling down event:
+```
+kubectl describe fleetautoscaler webhook-fleet-autoscaler
+```
+You should see these lines in events:
+```
+  Normal   AutoScalingFleet  11m                fleetautoscaler-controller  Scaling fleet simple-udp from 2 to 4
+  Normal   AutoScalingFleet  1m                 fleetautoscaler-controller  Scaling fleet simple-udp from 4 to 2
+```
+
+And get gameservers command output:
+```
+kubectl get gs -o=custom-columns=NAME:.metadata.name,STATUS:.status.state,IP:.status.address,PORT:.status.ports[0].port -n default
+```
+
+```
+NAME                     STATUS      IP               PORT
+simple-udp-884fg-6q5sk   Ready       35.247.117.202   7373
+simple-udp-884fg-b7l58   Allocated   35.247.117.202   7766
+```
+
+## Next Steps
+
+Read the advanced [Scheduling and Autoscaling](scheduling_autoscaling.md) guide, for more details on autoscaling.
+
+If you want to use your own GameServer container make sure you have properly integrated the [Agones SDK](../sdks/).
diff --git a/docs/fleetautoscaler_spec.md b/docs/fleetautoscaler_spec.md
@@ -21,6 +21,24 @@ spec:
       maxReplicas: 20
 ```
 
+Or for Webhook FleetAutoscaler below and in [example folder](../examples/webhookfleetautoscaler.yaml):
+
+```yaml
+apiVersion: "stable.agones.dev/v1alpha1"
+kind: FleetAutoscaler
+metadata:
+  name: fleet-autoscaler-example
+spec:
+
+  fleetName: fleet-example
+  policy:
+    type: Webhook
+    webhook:
+      name: "fleet-autoscaler-webhook"
+      namespace: "default"
+      path: "/scale"
+```
+
 Since Agones defines a new 
 [Custom Resources Definition (CRD)](https://kubernetes.io/docs/concepts/api-extension/custom-resources/) 
 we can define a new resource using the kind `FleetAutoscaler` with the custom group `stable.agones.dev` and API 
@@ -31,12 +49,78 @@ The `spec` field is the actual `FleetAutoscaler` specification and it is compose
 - `fleetName` is name of the fleet to attach to and control. Must be an existing `Fleet` in the same namespace
    as this `FleetAutoscaler`.
 - `policy` is the autoscaling policy
-  - `type` is type of the policy. For now, only "Buffer" is available
-  - `buffer` parameters of the buffer policy
+  - `type` is type of the policy. "Buffer" and "Webhook" are available
+  - `buffer` parameters of the buffer policy type
     - `bufferSize`  is the size of a buffer of "ready" game server instances
                     The FleetAutoscaler will scale the fleet up and down trying to maintain this buffer, 
                     as instances are being allocated or terminated
                     it can be specified either in absolute (i.e. 5) or percentage format (i.e. 5%)
     - `minReplicas` is the minimum fleet size to be set by this FleetAutoscaler. 
                     if not specified, the minimum fleet size will be bufferSize
-    - `maxReplicas` is the maximum fleet size that can be set by this FleetAutoscaler. Required. 
+    - `maxReplicas` is the maximum fleet size that can be set by this FleetAutoscaler. Required. 
+  - `webhook` parameters of the webhook policy type
+    - `service` parameters
+      - `name`  is the service name bound to Deployment of autoscaler webhook [(see example)](../examples/autoscaler-webhook/autoscaler-service.yaml)
+                      The FleetAutoscaler will scale the fleet up and down based on the response from this webhook server
+      - `namespace` is the kubernetes namespace where webhook is deployed. 
+                      If not specified, the "default" would be used
+      - `path` is the endpoint on the service  (i. e. /autoscale)
+    - `url` is a general URL to make post request to it (optional, instead of service)
+
+Note: only one `buffer` or `webhook` could be defined for FleetAutoscaler depends on `type` field. For Webhook policy we can define either `service` or `url`.
+
+
+# Webhook Endpoint Specification
+
+Webhook endpoint is used to delegate the scaling logic to a separate pod or server.
+
+FleetAutoscaler would send a request to the webhook endpoint every sync period (which is currently 30s) with a JSON body, and scale the target fleet based on the data that is returned.
+JSON payload with a FleetAutoscaleReview data structure would be sent to webhook endpoint and received from it with FleetAutoscaleResponse field populated. FleetAutoscaleResponse contains target Replica count which would trigger scaling of the fleet according to it.
+
+The connection to this webhook endpoint should be defined in `FleetAutoscaler` using Webhook policy type.
+
+```go
+type FleetAutoscaleRequest struct {
+	// UID is an identifier for the individual request/response. It allows us to distinguish instances of requests which are
+	// otherwise identical (parallel requests, requests when earlier requests did not modify etc)
+	// The UID is meant to track the round trip (request/response) between the Autoscaler and the WebHook, not the user request.
+	// It is suitable for correlating log entries between the webhook and apiserver, for either auditing or debugging.
+	UID types.UID `json:"uid""`
+	// Name is the name of the Fleet being scaled
+	Name string `json:"name"`
+	// Namespace is the namespace associated with the request (if any).
+	Namespace string `json:"namespace"`
+	// The Fleet's status values
+	Status v1alpha1.FleetStatus `json:"status"`
+}
+
+type FleetAutoscaleResponse struct {
+	// UID is an identifier for the individual request/response.
+	// This should be copied over from the corresponding FleetAutoscaleRequest.
+	UID types.UID `json:"uid"`
+	// Set to false if no scaling should occur to the Fleet
+	Scale bool `json:"scale"`
+	// The targeted replica count
+	Replicas int32 `json:"replicas"`
+}
+
+// FleetAutoscaleReview is passed to the webhook with a populated Request value,
+// and then returned with a populated Response.
+type FleetAutoscaleReview struct {
+	Request  *FleetAutoscaleRequest  `json:"request"`
+	Response *FleetAutoscaleResponse `json:"response"`
+}
+
+// FleetStatus is the status of a Fleet
+type FleetStatus struct {
+	// Replicas the total number of current GameServer replicas
+	Replicas int32 `json:"replicas"`
+	// ReadyReplicas are the number of Ready GameServer replicas
+	ReadyReplicas int32 `json:"readyReplicas"`
+	// AllocatedReplicas are the number of Allocated GameServer replicas
+	AllocatedReplicas int32 `json:"allocatedReplicas"`
+}
+```
+
+The example of the webhook written in Go could be found here [Example](../examples/autoscaler-webhook/main.go)
+It implements the scaling up logic based on the percentage of allocated gameservers in a fleet. If this fraction is more than threshold (i. e. 0.7) than `Replica` value is scaled by the factor (in this example twice) which results in creating more Ready GameServers.