rayyounghong/perforce-helix-core repository overview

⁠docker-perforce-helix-core

docker-perforce-helix-core is a Docker image for running Perforce Helix Core Server. Multi-arch: linux/amd64, linux/arm64.

mkdir -p ./perforce
docker run -d --name p4d \
  -p 1666:1666 \
  -e P4PORT=1666 \
  -v "$(pwd)/perforce:/perforce" \
  rayyounghong/perforce-helix-core:latest

The container runs p4d as the non-root perforce user. On first start it initializes the database under P4ROOT and writes server.id if missing.

⁠Docker Compose

services:
  p4d:
    image: rayyounghong/perforce-helix-core:latest
    container_name: p4d
    ports:
      - "1666:1666"
    environment:
      - P4PORT=1666
    volumes:
      - ./perforce:/perforce
    restart: unless-stopped

⁠Environment variables

⁠Volumes

/perforce holds all server state (metadata/ and logs/). Mount it as a named volume or bind mount to persist data across container restarts.

⁠Healthcheck

The image ships a healthcheck that runs p4 -p localhost:$P4PORT info every 30 seconds. Use docker ps or docker inspect to see the health status.

⁠Helix Proxy (p4p)

The image also ships p4p. Start it with the start-p4p entrypoint:

docker run -d --name p4p -p 1667:1666 \
  -e P4TARGET=p4d.example.com:1666 \
  -v "$(pwd)/p4p:/perforce" \
  --entrypoint /usr/local/bin/start-p4p \
  rayyounghong/perforce-helix-core:latest

⁠Local proxy lab

examples/proxy⁠ runs a p4d with simulated WAN latency and bandwidth (via tc netem), three independent p4p proxies in front of it (each with its own cache, like three remote sites), and a benchmark that compares p4 sync direct vs. cold vs. warm cache on every proxy:

cd examples/proxy
docker compose up -d --build                    # WAN_DELAY=50ms WAN_RATE=20mbit by default
./bench.sh                                      # FILES=200 SIZE_KB=256 by default
PROXIES=p4p2 ./bench.sh                         # benchmark only one proxy
WAN_DELAY=150ms WAN_RATE=5mbit docker compose up -d wan --force-recreate  # change link
docker compose down -v                          # tear down and delete data

⁠Monitoring traffic with Grafana

An opt-in monitoring profile adds cAdvisor (per-container network/CPU/memory counters), Prometheus (5s scrape) and Grafana with a pre-provisioned dashboard:

docker compose --profile monitoring up -d
./bench.sh

Read the dashboard by comparing who sends the bytes: on a direct sync p4d egress carries the files (capped at WAN_RATE); on a cold-cache proxy sync both p4d and p4p send them; on a warm-cache sync only p4p does, and p4d stays near zero. cAdvisor v0.60+ is required on Docker 29 (containerd image store). The wan container shares p4d's network namespace, so its counters duplicate p4d's.

From the host, point clients at localhost:1666 (direct) or localhost:1667 to 1669 (proxies). Fresh r26.1 servers default to security=4, so the first user must set a password (p4 passwd) and p4 login against each port.