epicmorg/p4d repository overview

⁠epicmorg/p4d

The Perforce Helix server (p4d) as a container: built FROM epicmorg/perforce:<release>⁠ on Debian 13 trixie, one tag per Helix release. The entrypoint stops the server the way Perforce recommends: on docker stop it runs p4 admin stop as a super user, so the journal and db.* files are closed cleanly instead of the process being killed.

⁠What's inside

• everything from epicmorg/perforce:<release> (p4d, p4, p4broker, p4dctl, ... of that release);
• entrypoint (under tini) that runs p4d -r ${P4ROOT} -J ${P4JOURNAL} [-L ${P4LOG}] [-A ${P4AUDITLOG}] [-C1] -p ${P4PORT} [-In ${P4NAME}] [-Id ${P4DESCRIPTION}] ${P4ARGS} <docker run args>;
• graceful stop on SIGTERM / SIGINT: optional p4 login (password from P4ADMIN_PASSWD or P4ADMIN_PASSWD_FILE), p4 admin stop, wait up to P4STOP_TIMEOUT seconds; if the stop is refused or times out, p4d gets a SIGTERM (the old behaviour);
• p4d-checkpoint - online checkpoint for docker exec, p4d-upgrade - offline db upgrade (p4d -xu) for a one-off container, like pg_upgrade (see below);
• volumes /perforce/root, /perforce/journals, /perforce/logs, /perforce/archives, /perforce/backups; port 1666.

Without P4ADMIN_USER / password the helpers still work on a server without protections, or when a valid ticket is available (P4TICKETS, ~/.p4tickets in a volume).

⁠Tags

Every tag is pushed to docker.io, quay.io and ghcr.io (epicmorg/p4d:<tag> on each) - same digest everywhere.

⁠Usage

services:
  p4d:
    image: epicmorg/p4d:r25.1
    restart: unless-stopped
    stop_grace_period: 3m          # longer than P4STOP_TIMEOUT, or docker kills p4d after 10 s
    ports:
      - "1666:1666"
    environment:
      P4NAME: p4main
      P4DESCRIPTION: Main P4 server
      P4CASE: "1"
      P4CHARSET: utf8
      P4AUDITLOG: /perforce/logs/auditlog
      P4ADMIN_USER: root
      P4ADMIN_PASSWD_FILE: /run/secrets/p4admin
      P4BACKUP_KEEP: "14"
    volumes:
      - ./root:/perforce/root
      - ./journals:/perforce/journals
      - ./logs:/perforce/logs
      - ./archives:/perforce/archives
      - ./backups:/perforce/backups
    secrets:
      - p4admin
secrets:
  p4admin:
    file: ./p4admin.txt

Checkpoint (online, server keeps running; journal rotated as usual) - e.g. from cron on the host:

docker exec p4d p4d-checkpoint      # -> /perforce/backups/<P4NAME>-<date>/<P4NAME>.ckp.N.gz (+ .md5)

Upgrade to a new release (offline, like pg_upgrade): stop the server, run the new image once with the same volumes and environment but without the entrypoint, then start the server on the new tag:

docker compose stop p4d
docker run --rm -it --entrypoint p4d-upgrade --env-file p4d.env \
  -v ./root:/perforce/root -v ./journals:/perforce/journals -v ./logs:/perforce/logs -v ./backups:/perforce/backups \
  epicmorg/p4d:r25.2            # offline checkpoint into /perforce/backups, then p4d -xu (--no-checkpoint to skip)
# set image: epicmorg/p4d:r25.2 in compose, then
docker compose up -d p4d

⁠Notes

• The default Docker stop timeout is 10 s - always raise it (stop_grace_period / --stop-timeout).
• p4 admin stop / admin checkpoint need a super user; with security level >= 3 that means a password (or a ticket).
• Restore = p4d -r /perforce/root -jr -z <checkpoint> in a one-off container (--entrypoint p4d) on an empty root.

⁠Links

• Source: EpicMorg/docker⁠ - issues and PRs welcome
• Tags float and are rebuilt weekly; pin by digest for reproducible builds: Tags, pinning and updates⁠
• CHANGELOG⁠ - License: MIT