Observability
Data-Driven Cluster Management
Instead of hopping between nodes over SSH, see the real state of your cluster on one screen with Prometheus-based metrics and a unified event stream.
Real-time metrics
CPU, RAM, disk, GPU and InfiniBand metrics are collected automatically from every node; in the job detail view, Slurm accounting data joins Prometheus series.
Alerting and event management
Rule-based alerts arrive via SMTP, webhook or in-panel notifications; silence windows and an ack flow keep alert fatigue away.
For Researchers
Submitting Jobs Has Never Been Easier
Instead of wrestling with sbatch scripts, researchers submit jobs from a web form, watch the queue and track their own resource usage.
Parameterized job templates
Turn frequent jobs into templates and share them with the team — users submit correctly configured jobs in seconds by filling in a few parameters.
Per-user permissions and attribution
Every job is submitted under the signed-in user's identity; fairshare, accounting and RBAC scoping — team leads see only their own team's jobs.
Enterprise Resilience
Runs Smoothly Even Air-Gapped
Offline environment support
Local RPM repositories, GPG-signed offline bundles and in-panel package import — complete installation and updates even on closed networks.
Security and auditability
Every mutation lands in the audit log automatically, with SHA-256 hash-chain tamper detection; nightly backups and config bundle export/import are standard.
Full Capabilities
Every Layer of the Cluster, Under One Roof
From job scheduling to hardware control, from provisioning to auditing — TULPAR's capability set covers the entire lifecycle of an HPC cluster.
Unified Cluster Dashboard
- Cluster health, workload and resource status on one screen
- Real-time, self-refreshing dashboards over WebSocket
- Node state distribution and live time series
- Quick-access cards for frequent operations
Slurm Job Management
- Submit, track, cancel/hold/release jobs — no CLI required
- Accurate fairshare and accounting with per-user attribution
- Partition, reservation, QOS and association management
- Self-service submission with parameterized job templates
PXE Provisioning & Image Management
- Node discovery and approval queue on Warewulf 4
- Automated OS installation with disk-install kickstart
- Image factory and versioned image deployment
- Node configurations converge declaratively via the agent
Infrastructure Service Management
- Install and manage NTP, LDAP, DNS and MySQL from the panel
- Config editors and one-click deploy
- Service health status and connection tests
- Live deployment logs
Identity & Access (RBAC)
- Hybrid local + LDAP/SSSD authentication
- Four-tier role model with scope-based permissions
- Team leads see only their own team's jobs
- JWT-secured sessions with automatic renewal
Real-Time Monitoring
- Prometheus-based CPU/RAM/disk/GPU/InfiniBand metrics
- SSH-free collection — scales naturally to 100+ nodes
- Slurm accounting joined with Prometheus in job details
- Central log store with in-panel log search
Alerting & Event Management
- Rule-based alerts: firing → notification → ack flow
- SMTP, webhook and in-panel notification channels
- Time-boxed silences and maintenance windows
- Unified event timeline
Docker Pod Management
- Container lifecycle: start, stop, restart
- Live log streaming with xterm.js
- Instant container metrics over WebSocket
- Compose viewer
Automation Engine
- Metric threshold or health event → drain/reboot/power/notify
- Dry-run mode for side-effect-free trials
- Detailed execution trace for every run
- Self-healing configuration drift
Security & Audit
- Automatic audit logging for every mutation
- One-click tamper verification with SHA-256 hash chain
- Rate limiting, security headers and hardening
- Single-source secrets with rotation procedures
Backup & Continuity
- Automatic nightly backups: databases + Slurm state
- Config bundle export/import — with mandatory dry-run report
- Night watchman: drift, surface and backup freshness checks
- One-click vendor support bundle
Hardware & Network Management
- Power control via BMC and SOL console access
- Network/IPAM: subnet definitions and automatic IP allocation
- GPU inventory and DCGM metrics
- Topology view and NFS storage management
Frequently Asked Questions
Got questions? Everything you need to know about installation, licensing and scaling is here.
TULPAR is an enterprise product offered under a commercial license. Licensing is annual, based on your cluster scale and the support coverage you need; every license includes the full feature set — there is no feature gating. Contact us through the contact page for a quote.
An EL9-based head server is enough. tulpar-preflight.sh prepares the prerequisites, install.sh brings up the whole stack; you then enter the admin password, NTP, LDAP and SLURM settings in the web wizard. A typical installation completes in minutes.
TULPAR talks to Slurm over the slurmrestd REST API and falls back to the CLI automatically, providing a stable, version-independent interface. Head Slurm components run natively on the host (dnf/systemd) and share the same repository and version as the compute nodes.
Yes. RPM repositories travel with the product, the Slurm package repository can be imported as a GPG-signed offline bundle, and container images are loaded with docker load. Every component works on a closed network, including node installation, monitoring and central logging.