Waiting for Redis "Ready to accept connections" in CI before tests

redis-server prints "Ready to accept connections" only after it finishes loading, and connections before that line are refused. The reliable gate in CI is to poll redis-cli ping until it returns PONG, or to attach a health check to the service.

What this error means

The first test occasionally fails with "Connection refused" and passes on re-run. The Redis log shows the "Ready to accept connections" line landing a moment after the job started connecting.

redis-server

1:M ... * Ready to accept connections tcp
# but the test ran a moment earlier and got:
Could not connect to Redis at localhost:6379: Connection refused

Common causes

A race between service boot and the first command

The container is reported created before redis-server is listening, so an eager client wins the race and gets refused.

No readiness gate in the workflow

Without a health check or PING loop, the job proceeds the instant the container exists rather than when Redis is ready.

How to fix it

Poll redis-cli ping until PONG

Block the job until Redis answers, then continue. This is deterministic and fast.

Terminal

until redis-cli -h localhost -p 6379 ping | grep -q PONG; do
  echo "waiting for redis to be ready..."; sleep 1
done
echo "redis is ready"

Attach a service health check

Let the runner hold the job until the service is healthy so tests never see a refused socket.

.github/workflows/ci.yml

services:
  redis:
    image: redis:7
    ports: ['6379:6379']
    options: >-
      --health-cmd "redis-cli ping"
      --health-interval 5s
      --health-retries 10

How to prevent it

Never connect on the assumption that a created container is ready.
Use a service health check or a PING loop as a hard gate.
Keep the readiness gate in a dedicated step before tests.