Add PodDisruptionBudgets and topologySpreadConstraints support

charts/shopsys-app/templates/pdb-storefront.yaml

@@ -0,0 +1,18 @@
+{{/* Rendered only when the storefront is guaranteed to run with 2+ replicas — a PDB with
+     minAvailable: 1 on a single-replica Deployment would block node drains forever.
+     Autoscaling counts only with minReplicas > 1 (the HPA may scale down to minReplicas). */}}
+{{- $s := .Values.storefront }}
+{{- if and ($s.pdb).enabled (or (and $s.autoscaling.enabled (gt (int $s.autoscaling.minReplicas) 1)) (gt (int $s.replicas) 1)) }}
+apiVersion: policy/v1
+kind: PodDisruptionBudget
+metadata:
+  name: storefront
+  labels:
+    {{- include "shopsys.labels" $ | nindent 4 }}
+    app: storefront
+spec:
+  minAvailable: {{ $s.pdb.minAvailable | default 1 }}
+  selector:
+    matchLabels:
+      app: storefront
+{{- end }}

charts/shopsys-app/templates/pdb-webserver.yaml

@@ -0,0 +1,18 @@
+{{/* Rendered only when the webserver is guaranteed to run with 2+ replicas — a PDB with
+     minAvailable: 1 on a single-replica Deployment would block node drains forever.
+     Autoscaling counts only with minReplicas > 1 (the HPA may scale down to minReplicas). */}}
+{{- $w := .Values.webserver }}
+{{- if and ($w.pdb).enabled (or (and $w.autoscaling.enabled (gt (int $w.autoscaling.minReplicas) 1)) (gt (int $w.replicas) 1)) }}
+apiVersion: policy/v1
+kind: PodDisruptionBudget
+metadata:
+  name: webserver-php-fpm
+  labels:
+    {{- include "shopsys.labels" $ | nindent 4 }}
+    app: webserver-php-fpm
+spec:
+  minAvailable: {{ $w.pdb.minAvailable | default 1 }}
+  selector:
+    matchLabels:
+      app: webserver-php-fpm
+{{- end }}

charts/shopsys-app/tests/component_schema_test.yaml

@@ -53,6 +53,25 @@ tests:
           path: spec.template.spec.containers[0].env[-1].name
           value: FROM_FIELD
 
+  - it: propagates topologySpreadConstraints into the pod spec
+    template: templates/deployment-storefront.yaml
+    set:
+      storefront:
+        topologySpreadConstraints:
+          - maxSkew: 1
+            topologyKey: topology.kubernetes.io/zone
+            whenUnsatisfiable: ScheduleAnyway
+            labelSelector:
+              matchLabels:
+                app: storefront
+    asserts:
+      - equal:
+          path: spec.template.spec.topologySpreadConstraints[0].topologyKey
+          value: topology.kubernetes.io/zone
+      - equal:
+          path: spec.template.spec.topologySpreadConstraints[0].maxSkew
+          value: 1
+
   - it: allows probe overrides
     template: templates/deployment-storefront.yaml
     set:

charts/shopsys-app/tests/pdb_test.yaml

@@ -0,0 +1,86 @@
+suite: pod disruption budgets
+values:
+  - ./values/required.yaml
+templates:
+  - templates/pdb-webserver.yaml
+  - templates/pdb-storefront.yaml
+tests:
+  - it: renders no PDB for single-replica components without autoscaling
+    asserts:
+      - hasDocuments:
+          count: 0
+
+  - it: renders the PDB when autoscaling is enabled
+    template: templates/pdb-webserver.yaml
+    set:
+      webserver:
+        autoscaling:
+          enabled: true
+    asserts:
+      - hasDocuments:
+          count: 1
+      - equal:
+          path: spec.minAvailable
+          value: 1
+      - equal:
+          path: spec.selector.matchLabels.app
+          value: webserver-php-fpm
+
+  - it: renders no PDB when autoscaling can scale down to a single replica
+    template: templates/pdb-webserver.yaml
+    set:
+      webserver:
+        autoscaling:
+          enabled: true
+          minReplicas: 1
+    asserts:
+      - hasDocuments:
+          count: 0
+
+  - it: renders no PDB when the pdb map is nulled by an overlay
+    template: templates/pdb-storefront.yaml
+    set:
+      storefront:
+        autoscaling:
+          enabled: true
+        pdb: null
+    asserts:
+      - hasDocuments:
+          count: 0
+
+  - it: renders the PDB when running multiple fixed replicas
+    template: templates/pdb-storefront.yaml
+    set:
+      storefront:
+        replicas: 3
+    asserts:
+      - hasDocuments:
+          count: 1
+      - equal:
+          path: spec.selector.matchLabels.app
+          value: storefront
+
+  - it: honors a minAvailable override
+    template: templates/pdb-webserver.yaml
+    set:
+      webserver:
+        autoscaling:
+          enabled: true
+        pdb:
+          minAvailable: 2
+    asserts:
+      - equal:
+          path: spec.minAvailable
+          value: 2
+
+  - it: can be disabled explicitly
+    template: templates/pdb-webserver.yaml
+    set:
+      webserver:
+        autoscaling:
+          enabled: true
+        pdb:
+          enabled: false
+    asserts:
+      - hasDocuments:
+          count: 0

charts/shopsys-app/values.schema.json

@@ -100,6 +100,13 @@
         }
       }
     },
+    "webserver": {
+      "type": "object",
+      "properties": {
+        "pdb": { "$ref": "#/$defs/pdb" },
+        "topologySpreadConstraints": { "type": "array" }
+      }
+    },
     "storefront": {
       "type": "object",
       "properties": {
@@ -110,13 +117,16 @@
         "secretEnv": {
           "type": "object",
           "additionalProperties": { "type": "string" }
-        }
+        },
+        "pdb": { "$ref": "#/$defs/pdb" },
+        "topologySpreadConstraints": { "type": "array" }
       }
     },
     "cron": {
       "type": "object",
       "properties": {
         "enabled": { "type": "boolean" },
+        "topologySpreadConstraints": { "type": "array" },
         "instances": {
           "type": "array",
           "items": {
@@ -134,6 +144,12 @@
     "consumers": {
       "type": "object",
       "properties": {
+        "defaults": {
+          "type": "object",
+          "properties": {
+            "topologySpreadConstraints": { "type": "array" }
+          }
+        },
         "instances": {
           "type": "array",
           "items": {
@@ -162,5 +178,15 @@
       }
     },
     "extraManifests": { "type": "array" }
+  },
+  "$defs": {
+    "pdb": {
+      "type": ["object", "null"],
+      "additionalProperties": false,
+      "properties": {
+        "enabled": { "type": "boolean" },
+        "minAvailable": { "type": ["integer", "string"] }
+      }
+    }
   }
 }

charts/shopsys-app/values.yaml

@@ -2,7 +2,8 @@
 #
 # Every workload component (webserver, storefront, cron, consumers) exposes the same
 # standard keys: image {repository, tag, pullPolicy}, replicas, resources, podAnnotations,
-# podLabels, nodeSelector, tolerations, affinity, extraEnv, extraVolumes, extraVolumeMounts,
+# podLabels, nodeSelector, tolerations, affinity, topologySpreadConstraints, extraEnv,
+# extraVolumes, extraVolumeMounts,
 # livenessProbe/readinessProbe, securityContext/podSecurityContext, lifecycle,
 # terminationGracePeriodSeconds, priorityClassName. Webserver and storefront additionally
 # expose independent `autoscaling` sections.
@@ -96,6 +97,13 @@ webserver:
     minReplicas: 2
     maxReplicas: 3
     targetCPUUtilization: 120
+  # PodDisruptionBudget protecting against voluntary disruptions (node drains). Rendered
+  # only when the component is guaranteed to run 2+ replicas (autoscaling enabled with
+  # minReplicas > 1, or fixed replicas > 1) - a minAvailable: 1 budget on a single
+  # replica would block drains forever.
+  pdb:
+    enabled: true
+    minAvailable: 1
   progressDeadlineSeconds: 1500
   terminationGracePeriodSeconds: 120
   strategy:
@@ -189,6 +197,7 @@ webserver:
                   operator: In
                   values:
                     - redis
+  topologySpreadConstraints: []
   podSecurityContext: {}
   extraEnv: []
   extraVolumes: []
@@ -208,6 +217,10 @@ storefront:
     minReplicas: 2
     maxReplicas: 3
     targetCPUUtilization: 120
+  # See webserver.pdb - same gating (autoscaling with minReplicas > 1, or replicas > 1).
+  pdb:
+    enabled: true
+    minAvailable: 1
   terminationGracePeriodSeconds: 60
   resources:
     limits:
@@ -251,6 +264,7 @@ storefront:
                   operator: In
                   values:
                     - storefront
+  topologySpreadConstraints: []
   podSecurityContext: {}
   securityContext: {}
   extraEnv: []
@@ -302,6 +316,7 @@ cron:
                 operator: In
                 values:
                   - background
+  topologySpreadConstraints: []
   podSecurityContext: {}
   extraEnv: []
   extraVolumes: []
@@ -341,6 +356,7 @@ consumers:
                   operator: In
                   values:
                     - background
+    topologySpreadConstraints: []
     podSecurityContext: {}
     securityContext: {}
     extraEnv: []

charts/shopsys-common/templates/_helpers.tpl

@@ -264,6 +264,10 @@ tolerations:
 affinity:
 {{ toYaml . | indent 2 }}
 {{- end }}
+{{- with .component.topologySpreadConstraints }}
+topologySpreadConstraints:
+{{ toYaml . | indent 2 }}
+{{- end }}
 {{- with .component.podSecurityContext }}
 securityContext:
 {{ toYaml . | indent 2 }}

charts/shopsys-infra/values.schema.json

@@ -17,14 +17,16 @@
       "properties": {
         "enabled": { "type": "boolean" },
         "config": { "type": "string" },
-        "replicas": { "type": "integer", "minimum": 0 }
+        "replicas": { "type": "integer", "minimum": 0 },
+        "topologySpreadConstraints": { "type": "array" }
       }
     },
     "rabbitmq": {
       "type": "object",
       "properties": {
         "enabled": { "type": "boolean" },
         "replicas": { "type": "integer", "minimum": 0 },
+        "topologySpreadConstraints": { "type": "array" },
         "auth": {
           "type": "object",
           "properties": {

charts/shopsys-infra/values.yaml

@@ -77,6 +77,7 @@ redis:
   nodeSelector: {}
   tolerations: []
   affinity: {}
+  topologySpreadConstraints: []
   podSecurityContext: {}
   securityContext: {}
   extraEnv: []
@@ -126,6 +127,7 @@ rabbitmq:
                   operator: In
                   values:
                     - rabbitmq
+  topologySpreadConstraints: []
   podSecurityContext: {}
   securityContext: {}
   extraEnv: []

docs/migrating-from-shopsys-deployment.md

@@ -133,3 +133,11 @@ Intentional differences of the phase-1 rewrite; everything else is a 1:1 port.
     overcommit, and anything bursting past `1Gi` is now OOMKilled instead of merely
     evictable — check your crons'/consumers' peak memory usage before relying on the
     defaults, and set `resources: null` on a component to restore the legacy behavior.
+22. **PodDisruptionBudgets for webserver and storefront**: the legacy package had none — a
+    node drain could evict all replicas at once. The chart now renders a
+    `minAvailable: 1` PDB per component whenever it is guaranteed to run 2+ replicas
+    (autoscaling enabled with `minReplicas > 1`, or fixed `replicas > 1`); setups that can
+    run a single replica get no PDB (it would block drains). Opt out
+    via `webserver.pdb.enabled` / `storefront.pdb.enabled`. `topologySpreadConstraints`
+    is also available as a standard component key (empty by default — the legacy
+    anti-affinity defaults are kept untouched).

docs/values.md

@@ -25,13 +25,18 @@ Every workload component (`webserver`, `storefront`, `cron`, `consumers.defaults
 | `autoscaling: {enabled, minReplicas, maxReplicas, targetCPUUtilization}` | per-component HPA (webserver + storefront) |
 | `resources` | container resources |
 | `podAnnotations` / `podLabels` | extra pod metadata |
-| `nodeSelector` / `tolerations` / `affinity` / `priorityClassName` | scheduling |
+| `nodeSelector` / `tolerations` / `affinity` / `topologySpreadConstraints` / `priorityClassName` | scheduling |
 | `extraEnv` | extra env entries (raw list, supports `valueFrom`) |
 | `extraVolumes` / `extraVolumeMounts` | additional volumes |
 | `livenessProbe` / `readinessProbe` | probe overrides |
 | `securityContext` / `podSecurityContext` | security contexts |
 | `terminationGracePeriodSeconds`, `lifecycle` | shutdown behavior |
 
+`topologySpreadConstraints` is a raw passthrough — supply complete constraints. Prefer
+`whenUnsatisfiable: ScheduleAnyway` unless spread is a hard requirement (`DoNotSchedule`
+can leave pods Pending on small clusters), and add `matchLabelKeys: [pod-template-hash]`
+so the spread ignores old-ReplicaSet pods during rolling updates.
+
 ## Top-level structure
 
 ```yaml
@@ -74,8 +79,12 @@ app:                         # shared backend configuration
   adminUrl: admin
   s3Endpoint: ""
 
-webserver:                   # component (see standard keys) + phpFpm/nginx sub-containers
-storefront:                  # component + its own `env` / `secretEnv` (storefront-secret-env Secret)
+webserver:                   # component (see standard keys) + phpFpm/nginx sub-containers;
+                             #   `pdb: {enabled: true, minAvailable: 1}` - rendered only with
+                             #   guaranteed 2+ replicas (autoscaling with minReplicas > 1,
+                             #   or replicas > 1)
+storefront:                  # component + its own `env` / `secretEnv` (storefront-secret-env Secret);
+                             #   `pdb` - same as webserver
 cron:                        # component + `instances: [{name, schedule}]`;
                              #   default resources: requests 100m/300Mi, limits 1Gi memory
 consumers:                   # `defaults` + `instances: [{name, transports, replicas, ...}]`;