ci: keep docker-compose.yml as the controller; move the agent to its own file

Turning ci/jenkins/docker-compose.yml from the controller into the agent meant
that redeploying a Jenkins set up from the old file would replace the
controller with an agent. Keep that file as the controller instead: the same
jenkins service and jenkins_home volume, now on the stock Java 25 image (its
old build context, ci/jenkins/Dockerfile, is the agent image now). The agent
moves to docker-compose.agent.yml. This replaces the previous commit's warning.
This commit is contained in:
Alexa
2026-10-25 11:50:09 +02:00
parent 29aca0e88c
commit 9c5b574d6e
3 changed files with 70 additions and 61 deletions
+19 -33
View File
@@ -1,39 +1,25 @@
# DorkNet build agent on Coolify. Deploy as a Docker Compose application from
# this repository with base directory /ci/jenkins, compose location
# /docker-compose.yml and "Connect to predefined network" on. It needs no domain: it dials out to
# Jenkins over WebSocket on Docker's internal network. Set the variables in
# the resource's Environment Variables; see ci/jenkins/README.md.
# The Jenkins controller: stock Jenkins on Java 25. It runs no builds; the
# client builds run on the agent (docker-compose.agent.yml). On Coolify,
# deploy it as a Docker Compose application from this repository (base
# directory /ci/jenkins) or paste it into a Compose service, with "Connect to
# predefined network" on so the agent can reach it. See ci/jenkins/README.md.
#
# This file used to define the Jenkins controller itself. If your Jenkins was
# deployed from it, don't redeploy that application from this version: it
# would replace the controller with an agent. Move the controller to its own
# service first and deploy this as a NEW application (README: "Upgrading from
# the old setup").
# Earlier versions of this file built the controller from ci/jenkins/Dockerfile,
# which is now the agent's image. The service name and data volume are the
# same, so redeploying an existing Jenkins from this file keeps its jobs and
# credentials; its builds then wait for the agent.
services:
agent:
build:
context: .
dockerfile: Dockerfile
jenkins:
# Pin a version with e.g. jenkins/jenkins:2.587-jdk25. Keep the agent on
# the same Java major version (its image defaults to Java 25).
image: ${JENKINS_IMAGE:-jenkins/jenkins:lts-jdk25}
# A restart from the Jenkins UI exits with code 5 and relies on Docker to
# start it again.
restart: unless-stopped
environment:
# Jenkins as seen from inside Docker (e.g. http://<jenkins container
# name>:8080), not the public URL, which is behind Cloudflare Access.
# Both are needed at runtime only; keep them out of the image build so
# the node secret never reaches a build argument.
JENKINS_URL: ${JENKINS_URL}
JENKINS_AGENT_NAME: ${JENKINS_AGENT_NAME:-dorknet-builder}
JENKINS_SECRET: ${JENKINS_SECRET}
JENKINS_WEB_SOCKET: "true"
JENKINS_AGENT_WORKDIR: /home/jenkins/agent
# Where the stock builds live in the archive (restic repository folders
# on the FTP, plain text). Overridable per run.
ARCHIVE_PC_REPO: ${ARCHIVE_PC_REPO:-}
ARCHIVE_QUEST_REPO: ${ARCHIVE_QUEST_REPO:-}
ARCHIVE_FTP_TLS: ${ARCHIVE_FTP_TLS:-explicit}
expose:
- "8080"
volumes:
- dorknet_cache:/home/jenkins/dorknet-cache
- agent_work:/home/jenkins/agent
- jenkins_home:/var/jenkins_home
volumes:
dorknet_cache:
agent_work:
jenkins_home: