# Slurm-web gives a Slurm cluster a browser interface without replacing squeue > Slurm-web is an open source web dashboard for Slurm based HPC clusters, with a feature set aimed at cluster operators, from GPU utilisation and rack topology to QOS views, multi-cluster support and Prometheus integration. It runs as three separate pieces, an agent, a gateway and a container image, and is offered with community support or a commercial SLA. **rackslab/Slurm-web** — Open source web interface for Slurm HPC & AI clusters - Repository: https://github.com/rackslab/Slurm-web - Website: https://slurm-web.com - Stars: 610 · Forks: 129 - Language: Python - License: MIT - Published: 2026-10-22 · Updated: 2026-10-22 · Language: en - Canonical page: https://hysenlabs.com/projects/rackslab-slurm-web ## Monitoring rather than scheduling Slurm-web is an open source web dashboard for Slurm based HPC clusters. The readme frames Slurm itself as the workload manager for HPC clusters, with a command line interface, and then positions Slurm-web as the graphical layer on top of it, with views to track jobs, insights and visualisations for monitoring the status of supercomputers in an organisation from a browser on any device. That boundary is the thing to hold on to. This is a monitoring and visibility tool, not a job submission front end, and the value it adds is the one the command line cannot: the squeue output is a list, and a list does not tell you at a glance which of four hundred jobs are stuck, which nodes are down, or how the GPU allocation is being used. For a shared cluster the audience is wider than the people who log in to the cluster, since the same state is useful to a research group leader, a support engineer and a systems administrator, and giving them a read only browser view is a different proposition from giving them a login. ## The feature list is a list of what operators actually ask for The list of capabilities in the readme reads like a backlog of support tickets, which is a good sign. There is a dashboard with interactive charts of resource and job status, instant filtering and sorting of jobs, live job status updates, and coloured badges so job state is visible without reading a word. Beyond that, GPU resources utilisation monitoring sits alongside an advanced visualisation of node status with rack topology, which is the feature that tells you whether a problem is one node or a whole rack, and an intuitive visualisation of the account tree, of quality of service definitions and of advanced reservations, which is the part that speaks to how a cluster is actually charged and prioritised. There is dark mode, custom branding with colours, logos and a favicon, multi-cluster support, advanced role based access control, and transparent caching, which matters because a cluster dashboard that hammers the controller with queries is a dashboard that slows the cluster. The last item is integration with Prometheus to collect and chart time series metrics of Slurm, which is what turns the dashboard from a live view into something you can alert on. ## Three processes, and the extras tell you what each one needs The packaging metadata is the clearest description of the architecture, because the optional dependency groups map one to one onto the components. There is an agent extra, which pulls in ClusterShell for talking to the cluster, a Prometheus client for metrics, a web enabled build of RacksDB as an embedded store, redis and the requests library. There is a gateway extra, which brings the markdown library and the LDAP and OIDC authentication plugins of the RFL framework. And there is a container extra, which adds gunicorn, the production WSGI server. So the design is an agent that runs near Slurm and collects state, a gateway that serves the web interface and handles authentication, and a containerised deployment of the web tier. The shared dependencies are Flask for the web framework and aiohttp for asynchronous calls, on top of five modules from the RFL library covering authentication, core, logging, settings and web. That foundation library is the clearest sign of the project's maturity, since the shared settings, logging and web layers are the parts every deployment gets wrong. ## Authentication covers directories and single sign-on Two authentication methods are offered and both sit in the gateway extra of the package. The first is LDAP, with explicit support for Active Directory, which is what most university and research environments already run and what a cluster administrator wants to avoid duplicating. The second is single sign-on through OpenID Connect, with Keycloak and Authentik given as the examples, which is the modern path and the one that plays best with a modern identity provider. The role based access control sits on top of whichever you choose, and the advanced permissions management is a listed feature rather than an afterthought, which is what you would expect from a tool that exposes resource and reservation configuration to a wider group of people. The practical advice is to decide the authentication method before you install, because it changes which extra you need and therefore which artefact you deploy, and to remember that the LDAP plugin and the OIDC plugin are both in the gateway extra rather than in the base package. ## Packaging, entry point and the linting setup The packaging is plain setuptools with a static version, and the project script is a single console entry point named slurm-web, which maps to a class method in the package. The package itself is found by an include pattern, so the distribution is the slurmweb namespace. The declared Python floor is 3.6, which is a low number for a project on its seventh major version, and it is worth checking against what your cluster actually runs, since a dashboard that needs a newer interpreter than your login nodes is a dashboard nobody can deploy. The linting configuration is deliberately narrow, selecting only the pycodestyle and pyflakes rules from Ruff rather than the full rule set, which tells you the project is not chasing a style argument and is spending its effort elsewhere. The tree is organised the way you would expect for a split application, with a frontend directory, a lib directory, a conf directory, a containers directory, a dev directory for the development environment, a docs directory, a tests directory and a scripts area, and a pre-commit configuration at the top level. ## Community support, and a commercial option from the same team The support section is worth reading because it is unusually well specified for an open source project. There is a Matrix chat room for quick feedback, with a note that a Matrix account is required and that one can be created on any public provider. There is a GitHub Discussions category for questions that need more context and for deeper answers. And the readme is honest that this community support is best-effort with participation from the Rackslab team, and that people might not be available to answer. Alongside it, the vendor offers commercial support with a service level agreement and a minimal response time, covering everything from installation assistance to advanced bug resolution, which is the standard shape for infrastructure software that a lab or company runs on a critical path. The project itself is developed and maintained by Rackslab, described as a software editor of open source solutions for HPC administration, operations and management, and the status section states plainly that Slurm-web is considered stable and ready for production. ## Conclusion Slurm-web suits a cluster administrator who has to answer the question that the command line answers badly, which is what is happening across the whole machine right now, and for a shared cluster where several people need to see the same state without a shared account on the login node. It is the wrong tool if you want to submit and manage jobs, since the description of the project is monitoring rather than scheduling, and the wrong choice if you cannot run a long lived agent on the cluster, because the agent is what collects the state. Verify three things before you deploy it: that your Slurm version and cluster layout are covered by the compatibility documentation, that the authentication method you need is available in the build you install, since LDAP and OIDC live in the gateway extra, and which cluster the gateway points at when you have several, since multi-cluster support exists but the configuration is yours to get right. The licence is MIT, the project is stated to be stable and production ready, and the newest release is v7.0.0 from 2026-06-11. ## FAQ ### What is Slurm-web? It is an open source web dashboard for Slurm based HPC clusters, providing a graphical interface to track jobs, monitor resource and GPU utilisation, visualise node status and rack topology, and view the account tree, quality of service definitions and reservations from a browser on any device. ### How is Slurm-web deployed? The quickstart guide points to container images and system packages for most Linux distributions. The package is built with setuptools and installs a single console command called slurm-web, and the project ships an agent, a gateway and a containerised web tier as separate pieces. ### Which authentication methods does Slurm-web support? LDAP with Active Directory support, and single sign-on through OpenID Connect, with Keycloak and Authentik given as examples. Both plugins live in the gateway optional dependency group rather than in the base package. ### Is Slurm-web suitable for production use? The readme states that Slurm-web is considered stable and ready for production. It is MIT licensed, the newest release listed is v7.0.0 from 2026-06-11, and the last push was on 2026-10-16. Community support is best-effort, with a commercial support option offering a service level agreement. ## Sources - [License: MIT](https://github.com/rackslab/Slurm-web/blob/main/LICENSE) - [Project website](https://slurm-web.com) - [rackslab/Slurm-web on GitHub](https://github.com/rackslab/Slurm-web) - [README](https://github.com/rackslab/Slurm-web/blob/main/README.md) - [Releases](https://github.com/rackslab/Slurm-web/releases) --- Hysen Labs editorial analysis, written from the project's own repository and release notes. Cite the canonical page: https://hysenlabs.com/projects/rackslab-slurm-web