docker-compose.multinode.yml

# Three-node CockroachDB cluster + Redis, for failover and consistency experiments.
# The single-node docker-compose.yml remains the default for day-to-day dev; this file
# is a separate topology you select explicitly with `-f docker-compose.multinode.yml`.
#
# Host ports:  SQL 26257/26258/26259   Admin UI 8089/8090/8091   Redis 6379
# Image tags pinned to match docker-compose.yml for reproducible `docker compose pull`.
#
# Default replication factor for user ranges is 3, so with three nodes the cluster
# tolerates exactly one node loss (2/3 quorum). Killing two loses quorum — that is the
# expected result, not a bug, and is itself worth measuring.

name: dtq-multinode

services:
  crdb1:
    image: cockroachdb/cockroach:v24.3.4
    command: start --insecure --join=crdb1,crdb2,crdb3 --advertise-addr=crdb1
    volumes:
      - crdb1-data:/cockroach/cockroach-data
    ports:
      - "26257:26257"
      - "8089:8080"

  crdb2:
    image: cockroachdb/cockroach:v24.3.4
    command: start --insecure --join=crdb1,crdb2,crdb3 --advertise-addr=crdb2
    volumes:
      - crdb2-data:/cockroach/cockroach-data
    ports:
      - "26258:26257"
      - "8090:8080"

  crdb3:
    image: cockroachdb/cockroach:v24.3.4
    command: start --insecure --join=crdb1,crdb2,crdb3 --advertise-addr=crdb3
    volumes:
      - crdb3-data:/cockroach/cockroach-data
    ports:
      - "26259:26257"
      - "8091:8080"

  # Runs `cockroach init` once, then exits. Retries because the nodes need a moment to
  # start listening, and an uninitialized node cannot answer a healthcheck yet
  # (the cluster is not live until init succeeds) — so this polls instead.
  crdb-init:
    image: cockroachdb/cockroach:v24.3.4
    depends_on:
      - crdb1
      - crdb2
      - crdb3
    restart: "no"
    entrypoint: ["/bin/sh", "-ec"]
    command:
      - |
        for i in $$(seq 1 30); do
          if /cockroach/cockroach init --insecure --host=crdb1:26257 2>&1 | tee /dev/stderr | grep -qE 'success|already been initialized'; then
            echo "cluster initialized"
            exit 0
          fi
          sleep 2
        done
        echo "init failed after 60s" >&2
        exit 1

  redis:
    image: redis:7.4.2-alpine
    ports:
      - "6379:6379"
    healthcheck:
      test: ["CMD", "redis-cli", "ping"]
      interval: 3s
      timeout: 3s
      retries: 5

volumes:
  crdb1-data:
  crdb2-data:
  crdb3-data: