fix(experimental-ec2-pattern): suspend more processes during deployment (#2752)

* fix(experimental-ec2-pattern): suspend more processes during deployment

* fix(experimental-ec2-pattern): suspend HealthCheck too

* Mention HealthCheck in changelog

---------

Co-authored-by: Jorge Azevedo <jorge.azevedo@guardian.co.uk>

Author: jacobwinch

.changeset/clever-tigers-bet.md

@@ -0,0 +1,16 @@
+---
+"@guardian/cdk": patch
+---
+
+The new deployment mechanism (`GuEc2AppExperimental`) now suspends some additional ASG processes:
+
+`AZRebalance`
+`InstanceRefresh`
+`ReplaceUnhealthy`
+`ScheduledActions`
+`HealthCheck`
+
+https://docs.aws.amazon.com/autoscaling/ec2/userguide/as-suspend-resume-processes.html#process-types
+
+This follows a recommendation from AWS and should make deployments (and rollbacks) more reliable:
+https://repost.aws/knowledge-center/auto-scaling-group-rolling-updates

src/experimental/patterns/__snapshots__/ec2-app.test.ts.snap

@@ -196,6 +196,11 @@ exports[`The GuEc2AppExperimental pattern matches the snapshot 1`] = `
           "PauseTime": "PT3M",
           "SuspendProcesses": [
             "AlarmNotification",
+            "AZRebalance",
+            "HealthCheck",
+            "InstanceRefresh",
+            "ReplaceUnhealthy",
+            "ScheduledActions",
           ],
           "WaitOnResourceSignals": true,
         },

src/experimental/patterns/ec2-app.ts

@@ -285,12 +285,18 @@ export class GuRollingUpdatePolicyExperimental {
       minSuccessPercentage: 100,
       waitOnResourceSignals: true,
       /*
-      If a scale-in event fires during an `AutoScalingRollingUpdate` operation, the update could fail and rollback.
-      For this reason, we suspend the `AlarmNotification` process, else availability of a service cannot be guaranteed.
-      Consequently, services cannot scale-out during deployments.
-      If AWS ever supports suspending scale-out and scale-in independently, we should allow scale-out.
+      All of these processes can launch and/or terminate instances and if this happens during a deployment it might
+      cause the rolling update to fail: https://repost.aws/knowledge-center/auto-scaling-group-rolling-updates
+      The full list of processes can be found here: https://docs.aws.amazon.com/autoscaling/ec2/userguide/as-suspend-resume-processes.html#process-types
        */
-      suspendProcesses: [ScalingProcess.ALARM_NOTIFICATION],
+      suspendProcesses: [
+        ScalingProcess.ALARM_NOTIFICATION,
+        ScalingProcess.AZ_REBALANCE,
+        ScalingProcess.HEALTH_CHECK,
+        ScalingProcess.INSTANCE_REFRESH,
+        ScalingProcess.REPLACE_UNHEALTHY,
+        ScalingProcess.SCHEDULED_ACTIONS,
+      ],
       /*
       Note: this is increased via an Aspect which also takes healthcheck grace period into account.