Nomad Operations and Planning
Nomad clusters are hosted on dedicated servers in the FD.io lab and used to manage Docker container based executors for FD.io project CI jobs.
Contents
Physical Lab Infrastructure
- FD.io CSIT git repository keeps an up-to-date FD.io lab specification.
- Server naming convention is specified here.
Nomad Operational Status
- TBD - add description or link to Nomad architecture / configuration
- TBD - Add links to Nomad monitoring status / data
Jenkins Nomad Plugin Configuration
Proposed Jenkins Nomad Plugin Label to Docker Image Mapping
Jenkins Nomad Plugin Labels can only be created via LF Service Desk requests. In order to test CI jobs in the Jenkins Sandbox, a different label must exist which points to a sandbox Dockerhub repository to avoid disrupting operational jobs during docker image testing. This is the proposed nomenclature for the mapping of Jenkins Nomad Plugin labels and the corresponding dockerhub repositories (production and sandbox variants).
All new docker images will first be pushed to the Dockerhub associated 'sandbox repo' and all CI jobs utilizing the Jenkins Nomad Plugin Label will be verified in the Jenkins Sandbox, by modifying the appropriate JJB YAML files to point to the sandbox repo. Once all CI jobs have been verified, then the image will be pushed to the operational dockerhub repository associated with the label.
NOTE: the 'sandbox labels' should NEVER be merged into the ci-management repository.
fdiotools-prod-{arch}-{os} fdiotools-sandbox-{arch}-{os}
QUESTION 1: Is there a need for project specific images?
QUESTION 2: Is there a need for different size labels (small, medium, large) with different resource allocations (eg. CPU, Memory, etc)?
| Label | Dockerhub Repo |
|---|---|
| fdiotools-prod-amd-ubuntu1804 | https://hub.docker.com/repository/docker/fdiotools/prod_amd_ubuntu1804 |
| fdiotools-prod-arm-ubuntu1804 | https://hub.docker.com/repository/docker/fdiotools/prod_arm_ubuntu1804 |
| fdiotools-sandbox-amd-ubuntu1804 | https://hub.docker.com/repository/docker/fdiotools/sandbox_amd_ubuntu1804 |
| fdiotools-sandbox-arm-ubuntu1804 | https://hub.docker.com/repository/docker/fdiotools/sandbox_arm_ubuntu1804 |
Legacy Jenkins Label to Docker Image Mapping
These labels are currently defined (circa VPP 20.05) and will be removed once there are no references to them in the ci-management repo. Some of them are not being used today.
| Legacy Label(s) | Dockerhub Repo |
|---|---|
| ubuntu1604-us, ubuntu1604-s, ubuntu1604-m, ubuntu1604-l, ubuntu1604-hub-us, ubuntu1604-nus, | https://hub.docker.com/repository/docker/snergster/vpp-ubuntu16 |
| ubuntu1604arm-hub-us | https://hub.docker.com/repository/docker/snergster/vpp-arm-ubuntu16 |
| ubuntu1804-us, ubuntu1804-s, ubuntu1804-m, ubuntu1804-l, ubuntu1804-hub-us, vpp-csit-device, vpp-csit-ubuntu18, | https://hub.docker.com/repository/docker/snergster/vpp-ubuntu18 |
| ubuntu1804arm-hub-us, ubuntu1804arm-s, ubuntu1804arm-m, vpp-csit-arm-ubuntu18 | https://hub.docker.com/repository/docker/snergster/vpp-arm-ubuntu18 |
| centos7-us, centos7-s, centos7-m, centos7-l, centos7-hub-us, | https://hub.docker.com/repository/docker/snergster/vpp-centos |
| ubuntu2004-us | https://hub.docker.com/repository/docker/snergster/vpp-ubuntu20 |
| centos8-us | https://hub.docker.com/repository/docker/snergster/vpp-centos8 |
Nomad Operations Tasks
This is the current list of high priority Nomad tasks.
| Task Description | Owner | % Complete | ETA |
|---|---|---|---|
| Move Nomad Docker images from https://hub.docker.com/search?q=snergster&type=image into fdiotools dockerhub account. | Dave W. | 25% | May 20, 2020 |
| Update Ubuntu1804 & Centos7 Nomad Docker images to include clang-9 toolchain packages required by VPP 'make install-deps'. | Dave W. | 10% | May 20, 2020 |
| Nomad server OS upgrades/normalization. Utilize ansible to create a uniform bare-metal OS environment across all Nomad servers. | Peter M. | 99.9% | May 29, 2020 |
| Build & test ubuntu 20.04 and centos8 Docker images for CI executors to run respective OS jobs. | Dave W. | ||
| Fix server-type-c4-3 (10.32.8.16) SDD with an HDD, reinstall Ubuntu 18.04 and restore to Nomad cluster. | Dave W. | TBD | Vexxhost Ticket Created |
| Update VPP ci-management configurations to use global jjb macros (lf-publisher & build-discarder) | Vanessa V. | ||
| Export Gerrit & Jenkins logs and other operational data to Nomad servers | Dave W. & Vanessa V. | TBD | LF Ticket Created |
Nomad Planning Wish List
This is the list of long term Nomad tasks. Please move them to the Nomad Operations Tasks and provide owner/ET information when they are being actively worked on.
- Add Nomad nodes to LF DNS & make the names the same as the hostname
- Add VPP 'make test-debug w/ ASAN enabled' verify job
- Convert Jenkins Nomad-plugin configuration spreadsheet to JJB managed YAML configuration files.
- Investigate Jenkins Nomad-plugin security issues.
- Convert Nomad/Jenkins/Gerrit monitoring/screen-scraping hacks into an operational monitoring system using exported gerrit & jenkins logs & nomad cli output.
- Add a mechanism to measure/track the memory consumed by the CI jobs inside Docker images. pmikus_comment: Depends if we want ability to do live monitoring or ability of storing logs (how long?). I can make prometheus to work for us by very simple change in config.
Completed Nomad Tasks
| Task Description | Owner | % Complete | Finish Date |
|---|---|---|---|
| Move Nomad build executor Dockerfiles from https://github.com/snergfdio/* into the ci-management project. | Dave W. | 100% | April 29, 2020 |
| Add a sudoer/admin account to all Nomad Servers. | Dave W. | 100% | May 18, 2020 |
| Move server-type-c4-2 from Class 's5ci' to Class 'builder' to cover t4-virl* nomad clients during upgrade. | Dave W. | 100% | May 18, 2020 |
| Perform fresh installation of Ubuntu 18.04 Server on t4-virl1, t4-virl2, & t4-virl3 | Peter M. | 100% | May 25, 2020 |
| Restore Nomad configuration on t4-virl1, t4-virl2, & t4-virl3 and rejoin on VPP cluster. | Peter M. | 100% | May 26, 2020 |
