deploy: zero-downtime rollout (2 replicas, start-first, healthchecks) for web & convex; document in runbook

docs/OPERACAO-PRODUCAO.md

@@ -97,6 +97,36 @@ Acesso
 - App: `https://tickets.esdrasrenan.com.br`
 - Convex: `https://convex.esdrasrenan.com.br` (o importante é o WebSocket do cliente conectar; o path `/version` responde para sanity‑check)
 
+## Zero‑downtime (sem queda durante deploy)
+
+Para evitar interrupção perceptível no deploy, habilitamos rollout "start-first" e múltiplas réplicas nos serviços web e Convex. O Traefik continua roteando para as tasks saudáveis enquanto uma task reinicia.
+
+O `stack.yml` já inclui:
+- `replicas: 2` para `web` e `convex_backend`.
+- `update_config.order: start-first` + `failure_action: rollback`.
+- `healthcheck` por porta local, garantindo que o Swarm só troque quando a nova task estiver OK.
+
+Se quiser ajustar recursos/estratégia:
+```
+deploy:
+  replicas: 2
+  update_config:
+    parallelism: 1
+    order: start-first
+    failure_action: rollback
+  restart_policy:
+    condition: any
+healthcheck:
+  # web
+  test: ["CMD", "node", "-e", "fetch('http://localhost:3000').then(r=>process.exit(r.ok?0:1)).catch(()=>process.exit(1))"]
+  interval: 10s
+  timeout: 3s
+  retries: 5
+  start_period: 30s
+```
+
+Observação: o CI já força `docker service update --force` após `stack deploy` e passa `RELEASE_SHA` no ambiente para variar a spec em todo commit, assegurando rollout.
+
 ### Dashboard (opcional)
 Você pode expor o painel do Convex para inspeção em produção.
 

stack.yml

@@ -26,7 +26,13 @@ services:
       RELEASE_SHA: "${RELEASE_SHA:-dev}"
     deploy:
       mode: replicated
-      replicas: 1
+      replicas: 2
+      update_config:
+        parallelism: 1
+        order: start-first
+        failure_action: rollback
+      restart_policy:
+        condition: any
       placement:
         constraints:
           - node.role == manager
@@ -40,6 +46,12 @@ services:
         - traefik.http.services.sistema_web.loadbalancer.server.port=3000
     networks:
       - traefik_public
+    healthcheck:
+      test: ["CMD", "node", "-e", "fetch('http://localhost:3000').then(r=>process.exit(r.ok?0:1)).catch(()=>process.exit(1))"]
+      interval: 10s
+      timeout: 3s
+      retries: 5
+      start_period: 30s
 
   convex_backend:
     image: ghcr.io/get-convex/convex-backend:latest
@@ -53,7 +65,13 @@ services:
       - CONVEX_SITE_ORIGIN=https://convex.esdrasrenan.com.br
     deploy:
       mode: replicated
-      replicas: 1
+      replicas: 2
+      update_config:
+        parallelism: 1
+        order: start-first
+        failure_action: rollback
+      restart_policy:
+        condition: any
       placement:
         constraints:
           - node.role == manager
@@ -67,6 +85,12 @@ services:
         - traefik.http.services.sistema_convex.loadbalancer.server.port=3210
     networks:
       - traefik_public
+    healthcheck:
+      test: ["CMD-SHELL", "curl -sf http://localhost:3210/version >/dev/null || exit 1"]
+      interval: 10s
+      timeout: 3s
+      retries: 5
+      start_period: 20s
 
   convex_dashboard:
     image: ghcr.io/get-convex/convex-dashboard:latest