fix(ci): update health check endpoint for review apps

- Return expected payload { version, timestamp, queue_services, data_services } in /health
- Fix infinite attempt loop in deploy check Github Action
- Untrack temporary wuzapi test scripts
This commit is contained in:
Rodrigo Borba 2026-02-26 16:35:15 -03:00
parent f2fb40afaa
commit 792951e4c8
3 changed files with 52 additions and 4 deletions

View File

@ -22,22 +22,23 @@ jobs:
run: echo "https://chatwoot-pr-${{ github.event.pull_request.number }}.herokuapp.com" run: echo "https://chatwoot-pr-${{ github.event.pull_request.number }}.herokuapp.com"
- name: Check Deployment Status - name: Check Deployment Status
run: | run: |
max_attempts=10 max_attempts=15
attempt=1 attempt=1
status_code=0 status_code=0
echo "Waiting for review app to be deployed/redeployed, trying in 10 minutes..." echo "Waiting for review app to be deployed/redeployed, trying in 10 minutes..."
sleep 600 sleep 600
while [ $attempt -le $max_attempts ]; do while [ $attempt -le $max_attempts ]; do
response=$(curl -s -o /dev/null -w "%{http_code}" https://chatwoot-pr-${{ github.event.pull_request.number }}.herokuapp.com/api) response=$(curl -s -o /dev/null -w "%{http_code}" https://chatwoot-pr-${{ github.event.pull_request.number }}.herokuapp.com/health)
status_code=$(echo $response | head -n 1) status_code=$(echo $response | head -n 1)
if [ $status_code -eq 200 ]; then if [ $status_code -eq 200 ]; then
body=$(curl -s https://chatwoot-pr-${{ github.event.pull_request.number }}.herokuapp.com/api) body=$(curl -s https://chatwoot-pr-${{ github.event.pull_request.number }}.herokuapp.com/health)
if echo "$body" | jq -e '.version and .timestamp and .queue_services == "ok" and .data_services == "ok"' > /dev/null; then if echo "$body" | jq -e '.version and .timestamp and .queue_services == "ok" and .data_services == "ok"' > /dev/null; then
echo "Deployment successful" echo "Deployment successful"
exit 0 exit 0
else else
echo "Deployment status unknown, retrying in 3 minutes..." echo "Deployment status unknown, retrying in 3 minutes..."
sleep 180 sleep 180
attempt=$((attempt + 1))
fi fi
else else
echo "Waiting for review app to be ready, retrying in 3 minutes..." echo "Waiting for review app to be ready, retrying in 3 minutes..."

View File

@ -2,6 +2,26 @@
# authentication, and callbacks. Used for health checks # authentication, and callbacks. Used for health checks
class HealthController < ActionController::Base # rubocop:disable Rails/ApplicationController class HealthController < ActionController::Base # rubocop:disable Rails/ApplicationController
def show def show
render json: { status: 'woot' } render json: {
version: Chatwoot.config[:version] || 'dev',
timestamp: Time.current.to_fs(:db),
queue_services: redis_status,
data_services: postgres_status
}
end
private
def redis_status
r = Redis.new(Redis::Config.app)
r.ping ? 'ok' : 'failing'
rescue StandardError
'failing'
end
def postgres_status
ActiveRecord::Base.connection.active? ? 'ok' : 'failing'
rescue StandardError
'failing'
end end
end end

View File

@ -0,0 +1,27 @@
# Correção do Deploy Check da Review App
## Objetivo
Corrigir a falha no pipeline de CI "Deploy Check" (`.github/workflows/deploy_check.yml`) que quebrava após 10 tentativas devido à Review App (Heroku) não retornar o healthcheck esperado.
## Contexto
O workflow do GitHub Actions esperava um JSON do endpoint `/api` contendo os campos `version`, `timestamp`, `queue_services` e `data_services` todos populados e com valor `"ok"` pros serviços. Porém, o endpoint `/api` (referente ao `ApiController#index`) não era exposto corretamente em alguns ambientes ou levantava erro 500 caso o Redis/Postgres demorassem a subir, além de cair em um loop infinito no script bash porque a variável `$attempt` não era incrementada se a chamada HTTP retornasse 200 mas o JSON fosse inválido.
## Passos Realizados
1. Mapeamos que já existia uma rota `get '/health', to: 'health#show'` apontando para o `HealthController` que apenas respondia `{ status: 'woot' }`.
2. Alteramos o `HealthController#show` para retornar o JSON robusto exigido pelo workflow, fazendo o ping no Redis e no Postgres e blindando as exceções com `rescue StandardError` para nunca retornar 500 durante a fase de boot.
3. Editamos o arquivo `.github/workflows/deploy_check.yml`:
- Trocamos o `curl` de `/api` para `/health`.
- Adicionamos a instrução `attempt=$((attempt + 1))` no bloco `else` (quando o teste do `jq` não passa), corrigindo o loop infinito.
- Aumentamos o `max_attempts` de 10 para 15 (dando 45 minutos de tolerância para a Review App subir o banco de dados e os dynos completamente).
## Principais Arquivos Alterados
- `app/controllers/health_controller.rb`
- `.github/workflows/deploy_check.yml`
## Como Validar
1. Subir essas alterações (commit e push) na branch do PR.
2. Acompanhar a aba "Actions" no GitHub e verificar o job "Check Deployment (pull_request)".
3. O script bash deverá fazer o cURL em `/health` e, assim que o PostgreSQL e Redis reportarem `"ok"`, o step será marcado como "Deployment successful".
## Como Reverter
Basta fazer um git revert do commit que adicionou essas alterações ou retornar os arquivos aos estados anteriores (o `HealthController` retornando apenas `{ status: 'woot' }` e o `deploy_check.yml` voltando para `/api` sem o incremento).