cc-backend

NOTE

While we do our best to keep the master branch in a usable state, there is no guarantee the master branch works. Please do not use it for production!

Please have a look at the Release Notes for breaking changes!

ClusterCockpit REST and GraphQL API backend

This is a Golang backend implementation for a REST and GraphQL API according to the ClusterCockpit specifications. It also includes a web interface for ClusterCockpit. This implementation replaces the previous PHP Symfony based ClusterCockpit web interface. The reasons for switching from PHP Symfony to a Golang based solution are explained here.

Overview

This is a Golang web backend for the ClusterCockpit job-specific performance monitoring framework. It provides a REST API and an optional NATS-based messaging API for integrating ClusterCockpit with an HPC cluster batch system and external analysis scripts. Data exchange between the web front-end and the back-end is based on a GraphQL API. The web frontend is also served by the backend using Svelte components. Layout and styling are based on Bootstrap 5 using Bootstrap Icons.

The backend uses SQLite 3 as the relational SQL database. While there are metric data backends for the InfluxDB and Prometheus time series databases, the only tested and supported setup is to use cc-metric-store as the metric data backend. Documentation on how to integrate ClusterCockpit with other time series databases will be added in the future.

For real-time integration with HPC systems, the backend can subscribe to NATS subjects to receive job start/stop events and node state updates, providing an alternative to REST API polling.

Completed batch jobs are stored in a file-based job archive according to this specification. The backend supports authentication via local accounts, an external LDAP directory, and JWT tokens. Authorization for APIs is implemented with JWT tokens created with public/private key encryption.

You find a detailed documentation on the ClusterCockpit Webpage.

Build requirements

ClusterCockpit requires a current version of the golang toolchain and node.js. You can check go.mod to see what is the current minimal golang version needed. Homebrew and Archlinux usually have current golang versions. For other Linux distros this often means that you have to install the golang compiler yourself. Fortunately, this is easy with golang. Since much of the functionality is based on the Go standard library, it is crucial for security and performance to use a current version of golang. In addition, an old golang toolchain may limit the supported versions of third-party packages.

How to try ClusterCockpit with a demo setup

We provide a shell script that downloads demo data and automatically starts the cc-backend. You will need wget, go, node, npm in your path to start the demo. The demo downloads 32MB of data (223MB on disk).

git clone https://github.com/ClusterCockpit/cc-backend.git
cd ./cc-backend
./startDemo.sh

You can also try the demo using the latest release binary. Create a folder and put the release binary cc-backend into this folder. Execute the following steps:

./cc-backend -init
vim config.json (Add a second cluster entry and name the clusters alex and fritz)
wget https://hpc-mover.rrze.uni-erlangen.de/HPC-Data/0x7b58aefb/eig7ahyo6fo2bais0ephuf2aitohv1ai/job-archive-demo.tar
tar xf job-archive-demo.tar
./cc-backend -init-db -add-user demo:admin:demo -loglevel info
./cc-backend -server -dev -loglevel info

You can access the web interface at http://localhost:8080. Credentials for login are demo:demo. Please note that some views do not work without a metric backend (e.g., the Analysis, Systems and Status views).

How to build and run

There is a Makefile to automate the build of cc-backend. The Makefile supports the following targets:

• make: Initialize var directory and build svelte frontend and backend binary. Note that there is no proper prerequisite handling. Any change of frontend source files will result in a complete rebuild.
• make clean: Clean go build cache and remove binary.
• make test: Run the tests that are also run in the GitHub workflow setup.

A common workflow for setting up cc-backend from scratch is:

git clone https://github.com/ClusterCockpit/cc-backend.git

# Build binary
cd ./cc-backend/
make

# EDIT THE .env FILE BEFORE YOU DEPLOY (Change the secrets)!
# If authentication is disabled, it can be empty.
cp configs/env-template.txt  .env
vim .env

cp configs/config.json .
vim config.json

#Optional: Link an existing job archive:
ln -s <your-existing-job-archive> ./var/job-archive

# This will first initialize the job.db database by traversing all
# `meta.json` files in the job-archive and add a new user.
./cc-backend -init-db -add-user <your-username>:admin:<your-password>

# Start a HTTP server (HTTPS can be enabled in the configuration, the default port is 8080).
# The --dev flag enables GraphQL Playground (http://localhost:8080/playground) and Swagger UI (http://localhost:8080/swagger).
./cc-backend -server  -dev

# Show other options:
./cc-backend -help

Project file structure

• .github/ GitHub Actions workflows and dependabot configuration for CI/CD.
• api/ contains the API schema files for the REST and GraphQL APIs. The REST API is documented in the OpenAPI 3.0 format in ./api/swagger.yaml. The GraphQL schema is in ./api/schema.graphqls.
• cmd/cc-backend contains the main application entry point and CLI implementation.
• configs/ contains documentation about configuration and command line options and required environment variables. Sample configuration files are provided.
• init/ contains an example of setting up systemd for production use.
• internal/ contains library source code that is not intended for use by others.
  • api REST API handlers and NATS integration
  • archiver Job archiving functionality
  • auth Authentication (local, LDAP, OIDC) and JWT token handling
  • config Configuration management and validation
  • graph GraphQL schema and resolvers
  • importer Job data import and database initialization
  • metricdispatch Dispatches metric data loading to appropriate backends
  • repository Database repository layer for jobs and metadata
  • routerConfig HTTP router configuration and middleware
  • tagger Job classification and application detection
  • taskmanager Background task management and scheduled jobs
  • metricstoreclient Client for cc-metric-store queries
• pkg/ contains Go packages that can be used by other projects.
  • archive Job archive backend implementations (filesystem, S3, SQLite)
  • metricstore In-memory metric data store with checkpointing and metric loading
• tools/ Additional command line helper tools.
  • archive-manager Commands for getting infos about an existing job archive, importing jobs between archive backends, and converting archives between JSON and Parquet formats.
  • archive-migration Tool for migrating job archives between formats.
  • convert-pem-pubkey Tool to convert external pubkey for use in cc-backend.
  • gen-keypair contains a small application to generate a compatible JWT keypair. You find documentation on how to use it here.
• web/ Server-side templates and frontend-related files:
  • frontend Svelte components and static assets for the frontend UI
  • templates Server-side Go templates, including monitoring views
• gqlgen.yml Configures the behaviour and generation of gqlgen.
• startDemo.sh is a shell script that sets up demo data, and builds and starts cc-backend.