Increase Cloud Deploy analysis step durations plus other QoL changes (#3230)

- Adjusts SLA analysis soak durations to 30 minutes per phase for sandbox and 1 hour per phase for production.
- Updates non-prod automations to auto-advance through canary-1 and canary-5 in crash and sandbox.
- Adds an automation rule to auto-promote releases from crash to sandbox (still need to manually approve the sandbox release).
- Adds documentation in skaffold.yaml clarifying dual deployment.

release/clouddeploy/delivery-pipeline.yaml

@@ -99,8 +99,8 @@ serialPipeline:
                   gcloud storage cp gs://${PROJECT_ID}-deploy/${TAG_NAME}/cloudbuild-schema-verify-${TARGET_ID}.yaml .
                   gcloud builds submit --no-source --config=cloudbuild-schema-verify-${TARGET_ID}.yaml
             analysis:
-              # 10 minutes.
+              # 30 minutes.
-              duration: 600s
+              duration: 1800s
               googleCloud:
                 alertPolicyChecks:
                     sandboxPartialDeploymentAlertPolicyChecks
@@ -108,8 +108,8 @@ serialPipeline:
             profiles: ["sandbox-partial-phase-5"]
             percentage: 50
             analysis:
-              # 10 minutes.
+              # 30 minutes.
-              duration: 600s
+              duration: 1800s
               googleCloud:
                 alertPolicyChecks:
                     sandboxPartialDeploymentAlertPolicyChecks
@@ -155,8 +155,8 @@ serialPipeline:
                   gcloud storage cp gs://${PROJECT_ID}-deploy/${TAG_NAME}/cloudbuild-deploy-gke-${TARGET_ID}.yaml .
                   gcloud builds submit --no-source --config=cloudbuild-deploy-gke-${TARGET_ID}.yaml
             analysis:
-              # 10 minutes.
+              # 30 minutes.
-              duration: 600s
+              duration: 1800s
               googleCloud:
                 alertPolicyChecks:
                     sandboxStableDeploymentAlertPolicyChecks
@@ -183,8 +183,8 @@ serialPipeline:
                   gcloud storage cp gs://${PROJECT_ID}-deploy/${TAG_NAME}/cloudbuild-schema-verify-${TARGET_ID}.yaml .
                   gcloud builds submit --no-source --config=cloudbuild-schema-verify-${TARGET_ID}.yaml
             analysis:
-              # 10 minutes.
+              # 1 hour.
-              duration: 600s
+              duration: 3600s
               googleCloud:
                 alertPolicyChecks:
                     productionPartialDeploymentAlertPolicyChecks
@@ -192,8 +192,8 @@ serialPipeline:
             profiles: ["production-partial-phase-5"]
             percentage: 50
             analysis:
-              # 10 minutes.
+              # 1 hour.
-              duration: 600s
+              duration: 3600s
               googleCloud:
                 alertPolicyChecks:
                     productionPartialDeploymentAlertPolicyChecks
@@ -250,8 +250,8 @@ serialPipeline:
                   gcloud storage cp gs://${PROJECT_ID}-deploy/${TAG_NAME}/cloudbuild-deploy-gke-${TARGET_ID}.yaml .
                   gcloud builds submit --no-source --config=cloudbuild-deploy-gke-${TARGET_ID}.yaml
             analysis:
-              # 10 minutes.
+              # 1 hour.
-              duration: 600s
+              duration: 3600s
               googleCloud:
                 alertPolicyChecks:
                     productionStableDeploymentAlertPolicyChecks
@@ -260,18 +260,34 @@ apiVersion: deploy.cloud.google.com/v1
 kind: Automation
 metadata:
   name: deploy-nomulus/auto-advance-canary
-description: Automatically advances rollouts through canary-1 phase after successful deployment and analysis.
+description: Automatically advances rollouts through canary-1 and canary-5 phases in non-prod.
 # Placeholder: Replace with project service account.
 serviceAccount: serviceAccount
 selector:
   targets:
   - id: crash
   - id: sandbox
-  - id: production
 rules:
 - advanceRolloutRule:
     id: advance-canary-phases
     sourcePhases:
     - "canary-1"
+    - "canary-5"
+    wait: 0m
+---
+apiVersion: deploy.cloud.google.com/v1
+kind: Automation
+metadata:
+  name: deploy-nomulus/auto-promote-crash-to-sandbox
+description: Automatically promotes release to sandbox after successful rollout in crash.
+# Placeholder: Replace with project service account.
+serviceAccount: serviceAccount
+selector:
+  targets:
+  - id: crash
+rules:
+- promoteReleaseRule:
+    id: promote-crash-to-sandbox
+    destinationTargetId: "sandbox"
     wait: 0m
 

release/clouddeploy/skaffold.yaml

@@ -5,6 +5,10 @@ metadata:
   name: nomulus-skaffold
 profiles:
 - name: crash
+  # Note: The stable profile intentionally includes partial-phase-5 manifests alongside
+  # primary workloads. This ensures that the candidate image is deployed across both deployments,
+  # maintaining stable pod and GKE node allocation to prevent churn on nodes hosting persistent
+  # registrar sessions.
   manifests:
     rawYaml:
     - ../../jetty/kubernetes/nomulus-crash-backend.yaml
@@ -40,6 +44,10 @@ profiles:
   deploy:
     kubectl: { }
 - name: sandbox
+  # Note: The stable profile intentionally includes partial-phase-5 manifests alongside
+  # primary workloads. This ensures that the candidate image is deployed across both deployments,
+  # maintaining stable pod and GKE node allocation to prevent churn on nodes hosting persistent
+  # registrar sessions.
   manifests:
     rawYaml:
     - ../../jetty/kubernetes/nomulus-sandbox-backend.yaml
@@ -75,6 +83,10 @@ profiles:
   deploy:
     kubectl: { }
 - name: production
+  # Note: The stable profile intentionally includes partial-phase-5 manifests alongside
+  # primary workloads. This ensures that the candidate image is deployed across both deployments,
+  # maintaining stable pod and GKE node allocation to prevent churn on nodes hosting persistent
+  # registrar sessions.
   manifests:
     rawYaml:
     - ../../jetty/kubernetes/nomulus-production-backend.yaml