Prologue -- If you're trying to build VyOS equuleus/sagitta ISO image with the usual way you may see following errors: ``` E: Failed to fetch http://dev.packages.vyos.net/repositories/equuleus/dists/equuleus/InRelease 403 Forbidden [IP: 104.18.30.79 443] E: The repository 'http://dev.packages.vyos.net/repositories/equuleus equuleus InRelease' is not signed. E: Failed to fetch http://dev.packages.vyos.net/repositories/sagitta/dists/sagitta/InRelease 403 Forbidden [IP: 104.18.30.79 443] E: The repository 'http://dev.packages.vyos.net/repositories/sagitta sagitta InRelease' is not signed. ``` You may also see `Sorry, you have been blocked` if you try to visit these links, but you aren't blocked - everyone is blocked. This is due to [change in VyOS policy](https://blog.vyos.io/community-contributors-userbase-and-lts-builds) where they don't offer their `dev.packages.vyos.net/repositories` for public anymore. This change applies only to stable branches (like 1.3 equuleus/1.4 sagitta), you can still build current/development branch with official repository. You want to continue to use VyOS long term? Then you can switch to current/development branch if you think that's good idea for your use case. If you like to use stable branch then you would need to obtain [VyOS subscription](https://vyos.io/subscriptions/support). The only other option currently is to build your own `dev.packages.vyos.net` package repository and that's what this project is all about. Purpose -- The goal of this project is to reproduce package repositories of stable branches formerly available at `dev.packages.vyos.net` and currently it's possible to use the automated scripts or the manual guide to reproduce the package repositories for **1.3.x equuleus** and **1.4.x sagitta**. The package repositories allow you to build *LTS* ISO with the usual slightly modified way. Host requirements and precautions -- All examples and scripts assume clean installation of **Debian 12 (Bookworm)**. Basic installation with `standard system utilities` is enough. We also recommend **dedicated virtual machine**. The build scripts are running under the `jenkins` user and thus in theory if you execute malicious build it can compromise your Jenkins and possibly your host. That's why you want dedicated OS and you don't want to share the Jenkins with other projects and ideally don't share the operating system with anything else either. This risk isn't likely - it would require compromised GitHub repositories to happen. The hardware requirements are significant: - 16GB total RAM (8GB RAM + 8GB swap is good option) - 100GB HDD - CPU will make builds faster or slower, there is no hard requirement The builds are memory hungry, but you don't need 16GB of physical RAM. You can have large swap to compensate, and you will still get good performance this way since the above 8GB threshold is reached only few times by few builds. Multiple options -- Historically this project documented the process how to build all packages and create package repository as manual guide with light usage of scripts. This is legacy method. Now we have possibility to do all steps from manual guide via automated scripts thanks to work of [@GurliGebis](https://github.com/GurliGebis). This is preferred method. We keep both methods in sync with changes. --- Option 1: Automated scripts == The automated scripts execute all manual steps with minimal interaction. The process is divided into 8 steps/scripts. These 8 scripts configure Jenkins, prepare the package repositories, build the packages and there is also one additional script to build ISO. **Obtain the scripts:** ```bash wget https://github.com/dd010101/vyos-jenkins/archive/refs/heads/master.tar.gz -O /tmp/vyos-jenkins.tar.gz tar -xf /tmp/vyos-jenkins.tar.gz -C /tmp mv /tmp/vyos-jenkins-master /opt/vyos-jenkins cd /opt/vyos-jenkins ``` **If you want to build only specific branch** Configure `BRANCH` environment variable to desired branch before you run any script. Not defined or empty value means all branches (the default). ```bash export BRANCH="sagitta" ``` **If you want to distribute ISO** Then you should remove VyOS branding, you can do this by configuring `NOT_VYOS` environment variable to `yes` before you run any script. Beware - by default, the ISO will include VyOS branding thus you shall not distribute the ISO. ```bash export NOT_VYOS="yes" ``` **Then execute each script and follow instructions:** - `1-prereqs.sh`- installs dependencies. - `2-jenkins.sh` - configures Jenkins, **interaction required**: - It asks you to **log-in into Jenkins**, after you do then confirm. - Then it asks you to **install recommended plugins in Jenkins**, after it's completed confirm. - Then it asks you to **create admin Jenkins account**, after you do then enter your username and confirm. - At last, it will ask you to **create Jenkins API Token**, after you do then you enter the token and confirm. - `3-repositories.sh` - creates empty package repositories. - `4-uncron.sh` - prepares uncron service. - `5-docker-jobs.sh` - builds vyos-build docker images, **takes a while**. - `6-provision-project-jobs.sh` - prepares package jobs in Jenkins. - `7-build-project-jobs.sh` - builds package jobs, **takes a long while**. - `8-nginx.sh` - configures nginx vhost for APT repositories. If all went well, then all steps should complete successfully and then you can: - `build-iso.sh` - builds the ISO :), **interaction required**, **takes a while**: - It asks you to specify branch equuleus or sagitta, after you do then confirm. - At least, it asks you to specify build-by, after you do then confirm and wait. This identifier is used as the `--build-by` parameter, this can be e-mail or any other identifier. And you should have the ISO(s) in current directory (`/opt/vyos-jenkins`). Jenkins will automatically detect changes and build new packages, thus if you keep build server running then it should keep the repositories up to date by itself. This way you can just use build-iso.sh again and again. You should check on the Jenkins **Build History** from time to time and/or before you build ISO to make sure all is going well. This is the same way how the official repository works. There is also option to shut down the OS and use it only when you need it. The Jenkins checks if any package needs rebuild in 1 hour interval, the check if 1 hour elapsed happens each 15th minute of hour. So if you boot the OS and start the Jenkins, then in worse case you would need to wait up to 15 minutes (to the closest 15th minute of hour), before rebuild of package would start. Then you shall wait before the **Build Queue** and **Build Executor Status** is empty, then make sure no build failed in the **Build History**, after this you can use build-iso.sh again. You could be also interested in the [Smoketest](#smoketest). If something isn't right, then see [Something is wrong](#something-is-wrong). --- Option 2: Manual guide == All following sections describe manual guide - if you use the automated scripts, then you don't need this information unless you face some issue, then this information is helpful to understand how things work and how to debug them. General expectations -- Unless specified otherwise all commands/scripts in the instructions should run as `root`. If you don't use root account then use `sudo -i` from your user to switch to root. Where other user is expected we provide note and `su` command. The current working directory doesn't matter unless specified with `cd`. The build system was designed to use 3 or more machines that's why some steps may seem a bit unusual. This guide merges everything to single host under single user to make it simpler and faster to get started. You may use another machine as build node for Jenkins (or multiple nodes), you may also use another machine for reprepro but here it's assumed everything is one host under one user. Before you install Jenkins, create its user and group -- Most of Jenkinfiles do respect your UID/GID but not all, for example [linux-kernel/Jenkinsfile](https://github.com/vyos/vyos-build/blob/equuleus/packages/linux-kernel/Jenkinsfile) has hardcoded UID and GID to 1006 and this will fail build if you don't have 1006:1006 user. That's why we want to create jenkins user and group with ID 1006 before installing Jenkins from apt. ```bash groupadd --system --gid 1006 jenkins useradd --system --comment Jenkins --shell /bin/bash --uid 1006 --gid 1006 --home-dir /var/lib/jenkins jenkins ``` If you have already existing user then please [change its UID/GID](extras/legacy-uid-gid.md). Install Jenkins, and its java -- Install java, then Jenkins. Let setup guide to install recommended plugins. Just follow the usual guide via APT https://www.jenkins.io/doc/book/installing/linux/#debianubuntu Install docker -- Just follow the usual guide via APT https://docs.docker.com/engine/install/debian/ Allow Jenkins to use docker: ```bash usermod -a -G docker jenkins ``` Setup local IP -- This guide will simplify the unknown by using static IP on dummy interface, this is hopefully outside your subnet if not please change all references of this IP with your own. ```bash cat << EOT >> /etc/network/interfaces auto dummy0 iface dummy0 inet static address 172.17.17.17/32 pre-up ip link add dummy0 type dummy EOT ``` ```bash ifup dummy0 ``` Now we can locally point to known IP `172.17.17.17` as it was the host itself. After adding docker group and/or after UID/GID change restart Jenkins -- ```bash systemctl restart jenkins.service ``` Launch local registry and set it, so it always runs when Docker runs -- ```bash docker run -d -p 5000:5000 --restart always --name registry registry:2.7 ``` **Allow insecure docker access to local registry:** Add your local IP with 5000 port to `insecure-registries` section in `/etc/docker/daemon.json`, something like this: ```bash cat << EOF > /etc/docker/daemon.json { "insecure-registries": [ "172.17.17.17:5000" ] } EOF ``` Then restart docker: ```bash systemctl restart docker.service ``` Install apt-cacher-ng for ELTS mirror -- This is currently used only by equuleus. ```bash apt install apt-cacher-ng ``` This will allow us to use `http://172.17.17.17:3142/deb.freexian.com/extended-lts` as ELTS mirror. Build patched vyos-build docker images -- The vyos/vyos-build docker image from dockerhub doesn't work for all packages as of now, thus we made some patches to make it work. If this changed in future then this step can be skipped. The below script clones the (patched) vyos-build, then builds and pushes the images to your custom Docker repository. ```bash #!/usr/bin/env bash set -e CUSTOM_DOCKER_REPO="172.17.17.17:5000" ELTS_MIRROR="http://172.17.17.17:3142/deb.freexian.com/extended-lts" # # Clone (patched) vyos-build git clone https://github.com/dd010101/vyos-build.git cd vyos-build/docker # # Build and Push equuleus git checkout equuleus docker build --build-arg "ELTS_MIRROR=$ELTS_MIRROR" \ --no-cache -t vyos/vyos-build:equuleus . docker tag vyos/vyos-build:equuleus ${CUSTOM_DOCKER_REPO}/vyos/vyos-build:equuleus docker push ${CUSTOM_DOCKER_REPO}/vyos/vyos-build:equuleus # # Build and Push sagitta git checkout sagitta docker build --no-cache -t vyos/vyos-build:sagitta . docker tag vyos/vyos-build:sagitta ${CUSTOM_DOCKER_REPO}/vyos/vyos-build:sagitta docker push ${CUSTOM_DOCKER_REPO}/vyos/vyos-build:sagitta # # Build and Push current -- (current is required for some sagitta packages) git checkout current docker build --no-cache -t vyos/vyos-build:current . docker tag vyos/vyos-build:current ${CUSTOM_DOCKER_REPO}/vyos/vyos-build:current docker push ${CUSTOM_DOCKER_REPO}/vyos/vyos-build:current ``` Install Jenkins plugins -- **Manage Jenkins -> Plugins -> Available plugins** - Docker - Docker Pipeline - Copy Artifact - SSH Agent - Pipeline Utility Steps - Job DSL Configure Built-In node -- **Manage Jenkins -> Nodes -> Built-In Node** **Add labels (tags)** - Docker - docker - ec2_amd64 Separated by space thus "Docker docker ec2_amd64" as result Configure Jenkins System -- **Manage Jenkins -> System** **Global properties -> Environmental Variables -> Add** > **Name:** DEV_PACKAGES_VYOS_NET_HOST\ > **Value:** jenkins@172.17.17.17 This user+IP/host will be used for SSH access to reprepro, it can be another host, we use the host itself, this IP needs to be accessible from docker container thus this should be LAN IP not localhost. **Global properties -> Environmental Variables -> Add** > **Name:** ARM64_BUILD_DISABLED\ > **Value:** true This is used to disable ARM64 support. The vyos-build expects that you have ARM64 build node and that's not something that is easy to obtain or emulate on x86. If you have ARM64 build node then skip this step and make sure your ARM64 node has tag `ec2_arm64`. If you try to build ARM64 without ARM node then most sagitta builds will wait and eventually fail. **Global properties -> Environmental Variables -> Add** > **Name:** CUSTOM_BUILD_CHECK_DISABLED\ > **Value:** true This is used to disable custom build check. Custom build check would normally skip upload to reprepro repository if package is built from non-vyos repository. Unfortunately currently it's impossible to build all packages from VyOS repositories, and thus we need to use custom repositories. Because some packages don't have functional build scripts or don't exist at all. This check doesn't make sense anyway since we are using our reprepro repository. **Global properties -> Environmental Variables -> Add** > **Name:** CUSTOM_DOCKER_REPO\ > **Value:** 172.17.17.17:5000 This variable is used to specify local docker registry for automatic `vyos-build` docker image rebuild, [see bellow for details](#additional-jobs). **Global Pipeline Libraries -> Add** > **Name:** vyos-build\ > **Project** repository: https://github.com/dd010101/vyos-build.git Currently patched version of vyos-build is required, in the future the official `https://github.com/vyos/vyos-build.git` may work but doesn't currently. Note for developers: equuleus is using only equuleus branch of vyos-build but sagitta is using both sagitta and current, thus if you fix something aimed at sagitta, you need to backports these changes to current as well, since some packages will use current and some sagitta branch. **Declarative Pipeline (Docker)** > **Docker registry URL:** http://172.17.17.17:5000 This is required to tell Jenkins to use your own (patched) vyos-build docker image and not the DockerHub version. Credentials for ssh-agent -- You need to set up SSH key authentication for the host specified in `DEV_PACKAGES_VYOS_NET_HOST` variable. Basically we want to allow Jenkins to SSH into itself with its own SSH key. Login as target user: ```bash su - jenkins ``` Generate regular SSH key: ```bash ssh-keygen -t ed25519 -C "jenkins" ``` Update authenticated_keys to allow Jenkins to log in to itself, something like this: ```bash cat ~/.ssh/id_ed25519.pub >> ~/.ssh/authorized_keys ``` Accept signature and verify SSH works: ```bash ssh 172.17.17.17 ``` Then you can add this private key to Jenkins: **Manage Jenkins -> Credentials -> System -> Global credentials (unrestricted) -> Add Credentials** > **Kind:** SSH Username with private key\ > **ID:** SSH-dev.packages.vyos.net\ > **Username:** jenkins **Private Key -> Enter directly -> Add** > Preparation for reprepro SSH host -- Install some packages: ```bash apt install reprepro gpg ``` Generate GPG singing key (without passphrase): ```bash sudo -u jenkins gpg --pinentry-mode loopback --full-gen-key ``` This key **needs to be without passphrase**. The reprepro uses this key in background thus there is no way to enter passphrase. Remember your pub key, it's random string like "934824D5C6A72DA964B3AFBD27A7E25D86BB7E2A". Create expected folder structure, prepare reprepro config and give Jenkins access, this is done for each release codename. Set SIGN_PUB_KEY: ```bash export SIGN_PUB_KEY="" ``` Set RELEASE name: ```bash export RELEASE=equuleus ``` or ```bash export RELEASE=sagitta ``` Then create reprepro repository for each RELEASE: ```bash export REPOSITORY=/home/sentrium/web/dev.packages.vyos.net/public_html/repositories/$RELEASE mkdir -p $REPOSITORY mkdir $REPOSITORY/conf cat << EOF > $REPOSITORY/conf/distributions Origin: $RELEASE Label: $RELEASE Codename: $RELEASE Architectures: source amd64 Components: main Description: $RELEASE SignWith: $SIGN_PUB_KEY EOF cat << EOF > $REPOSITORY/conf/options verbose EOF chown -R jenkins:jenkins /home/sentrium ``` uncron -- This is required addition for the reprepro. **Install dependencies** ```bash apt install opam ocaml socat ``` **Login as reprepro user and build uncon, then exit (if asked - confirm defaults)** You may have default opem switch already, then you will see `[ERROR] There already is an installed switch named default` - if you do then ignore this message and continue. ```bash su - jenkins git clone https://github.com/vyos/uncron.git cd uncron opam init opam switch create default 4.13.1 eval $(opam env --switch=default) opam install lwt lwt_ppx logs containers eval $(opam env) dune build exit ``` **Setup uncron service** ```bash cp /var/lib/jenkins/uncron/_build/install/default/bin/uncron /usr/local/sbin/ cat <<'EHLO' > /etc/systemd/system/uncron.service [Unit] Description=Command Queue Service After=auditd.service systemd-user-sessions.service time-sync.target [Service] EnvironmentFile=/etc/uncron.conf ExecStart=/usr/local/sbin/uncron ExecReload=/bin/kill -HUP $MAINPID KillMode=process User=jenkins Group=jenkins Restart=on-failure RestartSec=5s [Install] WantedBy=multi-user.target EHLO touch /etc/uncron.conf systemctl daemon-reload systemctl enable --now uncron.service chmod +x /var/lib/jenkins/uncron/src/uncron-add ``` **Create uncron-add script** We also use this as hack to fix some of VyOS packaging issues. ```bash cat << 'EOF' > /usr/local/bin/uncron-add #!/usr/bin/env bash set -e COMMAND="$1" # this is hack to workaround a issue where vyos didn't create sagitta branch # like the case of vyos-xe-guest-utilities, thus we need to build current # branch and then redirect from here to sagitta repository if [ ! -L ~/VyOS/current ]; then rm -rf ~/VyOS/current mkdir -p ~/VyOS/sagitta ln -s ~/VyOS/sagitta ~/VyOS/current fi if [[ "$COMMAND" == *"repositories/current"* ]]; then COMMAND=${COMMAND//current/sagitta} fi if [[ "$COMMAND" == *"vyos-xe-guest-utilities"* ]] && [[ "$COMMAND" == *"current"* ]]; then COMMAND=${COMMAND//current/sagitta} fi /var/lib/jenkins/uncron/src/uncron-add "$COMMAND" EOF chmod +x /usr/local/bin/uncron-add ``` Multibranch Pipelines (by script) -- Script exists to automate pipeline/job creation. **Generate Jenkins API token:** **User account (click on your username - top-right corner) -> Configure**\ **API Token -> Current token(s) -> Add new Token** > **Name:** CLI\ > -> [Generate] -> [copy generated key and save it for later use] **Get the script** ```bash git clone https://github.com/dd010101/vyos-jenkins.git cd vyos-jenkins/manual ``` **Install dependencies** ```bash apt install -y xmlstarlet jq ``` **Adjust settings to suit your Jenkins** ```bash export JENKINS_USER= export JENKINS_TOKEN= ``` **If you want to build only specific branch** Configure `BRANCH` environment variable to desired branch before you run the script. Not defined or empty value means all branches (the default). ```bash export BRANCH="sagitta" ``` **Create jobs** Then wait for branch indexing to complete. ```bash ./seed-jobs.sh create ``` **After branch indexing you can trigger build for everything** Make sure you have >=16GB RAM or 8GB RAM + 8GB swap, since running build for everything like this eats more memory than building one by one this is also dependent on how many Number of executors you have. ```bash ./seed-jobs.sh build ``` Now wait for build to complete and check Build History and Dashboard for failed builds. If you find any failed builds then read Console Output to see why it did failed. This process is required only once. After you create Jenkins jobs, and you do first build then Jenkins will periodically check if GIT repository for job changed and will do automatically build given job/packages. You can also create Multibranch Pipelines manually, [see bellow](#multibranch-pipelines-manual). Mirror preparation -- Use the default procedure to build ISO (via docker) but you need to specify your `--vyos-mirror` and your gpg singing key `--custom-apt-key`. To make `--vyos-mirror` is easy, you just install your favorite webserver and point the webroot to `/home/sentrium/web/dev.packages.vyos.net/public_html/repositories/`. For example nginx vhost looks something like this: ``` server { listen 80; listen [::]:80; root /home/sentrium/web/dev.packages.vyos.net/public_html/repositories; autoindex on; server_name _; location / { try_files $uri $uri/ =404; } location ~ /(.*)/conf { deny all; } location ~ /(.*)/db { deny all; } } ``` This will give you HTTP APT repository, like this `http://172.17.17.17/equuleus`. To create `--custom-apt-key` you need to export your gpg singing public key, for example: ```bash sudo -u jenkins gpg --armor --output /home/sentrium/web/dev.packages.vyos.net/public_html/repositories/apt.gpg.key \ --export-options export-minimal --export vyos ``` This will give you `/home/sentrium/web/dev.packages.vyos.net/public_html/repositories/apt.gpg.key` or `http://172.17.17.17/apt.gpg.key`. If you have the `--vyos-mirror` URL of your own repository and your own singing key `--custom-apt-key` then you can use these two pieces to build ISO from your own package mirror. How to build ISO -- Complete instructions to illustrate how to build ISO via Docker method included with the extra options outline above. We assume you already have Docker running if not follow the [official instructions](https://docs.docker.com/engine/install/debian/) for your OS. **Obtain the sources:** ```bash git clone https://github.com/dd010101/vyos-build cd vyos-build ``` **Select branch of your choice** For all following steps will use BRANCH environment variable since the branch repeats a lot. ```bash export BRANCH=equuleus ``` or ```bash export BRANCH=sagitta ``` **Switch to branch** ```bash git checkout "$BRANCH" ``` **Clear previous build resources (if any)** There is `make clean` but that doesn't always clean everything and may produce corrupted build environment. The `make clean` is trying to remove specific parts of `build` directory, but it doesn't always do so correctly. This happens mainly if you switch branches - that's why it's better to always delete the whole `build` directory. ```bash rm -rf build/ ``` **Obtain vyos-build docker container** You can reuse your docker container image if you already have [custom registry](#launch-local-registry-and-set-it-so-it-always-runs-when-docker-runs) with [patched vyos-build docker container](#build-patched-vyos-build-docker-images). Change the registry URL if you build on other machine. ```bash docker pull "172.17.17.17:5000/vyos/vyos-build:$BRANCH" docker tag "172.17.17.17:5000/vyos/vyos-build:$BRANCH" "vyos/vyos-build:$BRANCH" ``` If you don't have custom registry then build the container - this will take a while: ```bash docker build -t "vyos/vyos-build:$BRANCH" docker ``` You should rebuild the container from time to time - not very frequently but sometimes the build will break if you have too old container. **Obtain apt singing key for your custom mirror** ```bash wget http://172.17.17.17/apt.gpg.key -O /tmp/apt.gpg.key ``` **Launch the vyos-build docker container** This is the usual run command from official documentation, we need to add extra mount for our apt singing key for later use via `-v "/tmp/apt.gpg.key:/opt/apt.gpg.key"`. The docker run command will mount current working directory for use inside the container that's why you need to execute this command inside the `vyos-build` directory (that is the GIT repository you cloned above). You can also replace the `-v "$(pwd)":/vyos` with static path if you like not to depend on current directory (for example `-v /opt/vyos-build:/vyos`). ```bash docker run --rm -it \ -v "$(pwd)":/vyos \ -v "/tmp/apt.gpg.key:/opt/apt.gpg.key" \ -w /vyos --privileged --sysctl net.ipv6.conf.lo.disable_ipv6=0 \ -e GOSU_UID=$(id -u) -e GOSU_GID=$(id -g) \ "vyos/vyos-build:$BRANCH" bash ``` Now we should be inside the container. **Configure and build the ISO** Command for configuring changed over time, equuleus has `./configure`, sagitta has `./build-vyos-image iso` instead. You may want to customize the configuration options, see what is available: For equuleus: ```bash sudo ./configure --help ``` For sagitta: ```bash sudo ./build-vyos-image --help ``` We need to add extra two options to configure `--vyos-mirror` and `--custom-apt-key`. We also add smoketest via `--custom-package vyos-1x-smoketest` for good measure. Here are examples - please adjust options to your liking: For equuleus: ```bash sudo ./configure --architecture amd64 --build-by "myself@localhost" \ --build-type release --version "1.3.x" \ --vyos-mirror http://172.17.17.17/equuleus --custom-apt-key /opt/apt.gpg.key \ --debian-elts-mirror http://172.17.17.17:3142/deb.freexian.com/extended-lts \ --custom-package vyos-1x-smoketest \ && sudo make iso ``` For sagitta: ```bash sudo ./build-vyos-image iso --architecture amd64 --build-by "myself@localhost" \ --build-type release --version "1.4.x" \ --vyos-mirror http://172.17.17.17/sagitta --custom-apt-key /opt/apt.gpg.key \ --custom-package vyos-1x-smoketest ``` This will take a while - after all is done then you can `exit` the container and you should have `build/live-image-amd64.hybrid.iso`. Something is wrong -- You may face situation when Jenkins build may fail or doesn't produce .deb packages and thus ISO build fails with unmet dependencies. Sometimes the Jenkins build fails for temporary reason like network/server issue, thus simple retry with **Build now** will fix the failure. There are two logs you should check for pointers. 1) In Jenkins - find the job/packages of your interest - select branch of interest and find last run with `Git SHA1: ...`. There may be other runs without `Git SHA1: ...` - those aren't build runs, those are branch indexing runs that check if package needs rebuild - ignore those. If you don't see any runs then use the **Build now** action to trigger new build run. In specific run you should see **Console Output**. 2) The `uncron.service` has log file you can access via `journalctl --no-pager -b -u uncron.service`, look for package in question and check if there isn't error output or `Job exited with code 0` other than 0. If you face errors that don't make sense then it's likely your docker container is outdated or your Jenkins configuration is missing some pieces. Thus, as first thing you can try to rebuild the [docker container](#build-patched-vyos-build-docker-images) and if that doesn't help then verify all the Jenkins configuration is in its place. Smoketest -- It's not bad idea to sanity check your ISO image. This is why VyOS has testing procedure called smoketest. This test checks mainly configuration, it's not complete test, it will only tell you if something is very wrong, passed test doesn't mean image will be necessarily fully functional - with that said it's still useful, just don't put too much trust in it. You will need host that supports virtualization and thus can run KVM. Main test (`make test`) takes around two hours to complete, additional tests are significantly faster. There is way to run automated smoketest via `vyos-build/Jenkinsfile` but that requires Jenkins, and also it starts all tests in parallel thus requiring more RAM since multiple virtual machines run in parallel. This method doesn't allow selecting the test either. There is requirement to include `vyos-1x-smoketest` package in your ISO image build. By default, ISO build doesn't include smoketest thus you need to include it via the usual parameter for custom packages. When you run `./configure` (equuleus) or `./build-vyos-image` (sagitta) add `--custom-package vyos-1x-smoketest`. If you have your image with smoketest then you can test it. If you did build it yourself, then your likely already have clone of `vyos-build` repository. If not clone it and switch to your branch. ```bash git clone https://github.com/dd010101/vyos-build.git cd vyos-build # pick one git checkout sagitta git checkout equuleus ``` There is known issue with smoketest that it will fail if you have too many CPU cores/threads. The test is designed to use half of your cores, but it will fail if calculates more than 4, thus if you have 8 or more cores/threads then test likely will fail. If you do apply this patch to cap cores to 3: ```bash sed -i 's~cpu /= 2~cpu = 3~' scripts/check-qemu-install ``` More cores don't increase speed, the test is single thread anyway, it usually uses <2 cores and thus 3 is more than enough. More cores will not speed up the test - it will only make the test fail due to OOM inside the test virtual machine. If you did build your image, then you should have `build` directory and there should be your ISO as `live-image-amd64.hybrid.iso`. If you want to test arbitrary ISO then you can plant whatever ISO into the expected path `build/live-image-amd64.hybrid.iso`. You of course need ISO image that includes `vyos-1x-smoketest` package. Install dependencies: ```bash apt install qemu-kvm python3-tomli python3-pexpect ``` And then you can launch virtual machine to do tests thing via `make`. There are multiple tests: CLI configuration test ```bash make testd ``` There is also `make test` that runs identical tests to the `make testd` the difference is if the `vyos-configd.service` service is enabled or not and VyOS enables this service by default, that's why `make testd` is more accurate. Configuration file load test ```bash make testc ``` RAID1 test ```bash make testraid ``` You can as well run smoketest directly from **installed** VyOS, but you need many network interfaces otherwise some tests will fail. The test is expecting 8 or more network ports. You do it by simply including `vyos-1x-smoketest` package in your ISO image build. Then you can boot and run `vyos-smoketest`. I'm not sure if all 8 are required but with few the test will fail for sure. Also make sure you have <= 4 cores/threads and 4GB or more of RAM. If you have more cores/threads then you need more RAM as well, try to keep it 1:1. Smoketest eats a lot of RAM and more so if you have more cores/threads. Multibranch Pipelines (manual) -- Use + button on Jenkins dashboard to add Multibranch Pipeline. Each Jenkinsfile needs its own Multibranch Pipeline, the setup is the same for all packages, and you just adjust location of Jenkinsfile and/or GIT repository to whatever you want to build. See packages info bellow for list of all GIT repository and location their Jenkinsfile. It makes sense to configure one pipeline and then use "Copy from" and just change the Jenkinsfile location and/or GIT repository. Start with something small like dropbear, and after you verify your setup works then try the bigger stuff like linux-kernel. There are two types of configurations - first you have single shared GIT repository with many Jenkinsfile, like the https://github.com/vyos/vyos-build.git, thus each pipeline points to different Jenkinsfile, but they share same GIT url. Rest of the packages have their own GIT repository with single root Jenkinsfile, thus the Jenkinsfile stays "Jenkinsfile" and you change the GIT url only. The https://github.com/vyos/vyos-build.git repository has also its own root Jenkinsfile - ignore it since that one is trying to build ISO with default (blocked) apt mirror. You can see all packages and information about them in the [packages.md](extras/packages.md) file. **Branch Sources -> Add source -> Git** > **Project Repository:** https://github.com/vyos/vyos-build.git (or any other repository, like https://github.com/vyos/vyos-1x.git) You may want to restrict to only branches you care about, thus: **Behaviours -> Add -> Filter by name (with regular expression)** > *Regular expression:** (equuleus|sagitta) **Behaviours -> Add -> Advanced clone behaviours** > **Fetch tags:** [✓] (leave defaults) Advanced clone is required for some packages to obtain all tags (specifically `vyos-cloud-init` and `vyos-1x`). It doesn't hurt to have advanced clone for everything, thus you can set-it and copy for everything without worrying about what to use it for. **Build Configuration -> Mode (by Jenkinsfile)** > **Script Path:** packages/dropbear/Jenkinsfile (if you want to build package from vyos/vyos-build repository) > **Script Path:** Jenkinsfile (or leave just Jenkinsfile if you want to build repository like vyos/vyos-1x where there is just one package) **Scan Multibranch Pipeline Triggers** > [✓] **Periodically if not otherwise run** > > **Interval:** 1 hour Jenkins will check the source GIT repository if changes were made and execute automatic build if needed. This will keep packages up to date. Try to build -- Now it's possible to select some **Multibranch Pipeline**, select your **branch**, and you can press build button to see what happens! If all is well you should see .deb appearing in `/home/sentrium/web/dev.packages.vyos.net/public_html/repositories/`: ```bash find /home/sentrium/web/dev.packages.vyos.net/public_html/repositories/ -name '*.deb' -print ``` If build fails then click the specific build number and check **Console Output** for hints why it does so.