SLURM Manager

SSH-based HPC job monitor and controller for Android

SLURM Manager connects to one or more HPC clusters over SSH and lets you monitor, filter, and control Slurm jobs directly from your phone or tablet. No VPN client or web portal required — just an SSH key or password.

---

Screenshots

Phone

Tablet

Medium tablet

---

Features

• Multi-cluster support — add as many servers as you need; switch with a single filter chip
• Live job list — background polling via WorkManager keeps the job list fresh
• Job detail — resources, timeline, runtime, and one-tap cancel / requeue
• Job submission — pick a local sbatch script or type one inline, set partition / nodes / CPUs / time limit
• Alert rules — per-server threshold alerts for queue wait time, runtime exceeded, node failures, and partition down events
• Push notifications — state-change and threshold alerts delivered as Android notifications (with progress indicators for running jobs)
• Adaptive layout — two-pane list-detail on tablets and large phones; bottom nav on compact; rail on medium
• Material You — full dynamic color (Android 12+) with expressive spring motion
• SSH auth — password or private key (Ed25519, RSA, ECDSA); private keys stored in EncryptedSharedPreferences

---

Architecture

app/src/main/java/com/tomasps/slurmmanager/
├── data/
│   ├── credential/      EncryptedSharedPreferences key store
│   ├── local/db/        Room database — Job + Server entities, DAOs
│   ├── notification/    NotificationEngine (state-change + alert notifications)
│   ├── remote/ssh/      SshClient (sshj), SlurmOutputParser (squeue output → domain models)
│   ├── repository/      JobRepositoryImpl, ServerRepositoryImpl
│   └── worker/          PollWorker (WorkManager), AlertEngine, JobDiffer
├── di/                  Hilt modules (AppModule, DataStoreModule, WorkerModule)
├── domain/
│   ├── model/           Job, Server, JobState, AlertRule, AlertType, AuthMethod, …
│   └── repository/      JobRepository, ServerRepository interfaces
└── presentation/
    ├── dashboard/        Job list with server filter chips, FAB, pull-to-refresh
    ├── jobdetail/        Resources, timeline, cancel/requeue actions
    ├── onboarding/       Add-server wizard with connection test
    ├── serverdetail/     Per-server jobs, history, polling/alert settings
    ├── servers/          Server list with swipe-to-delete
    ├── settings/         Global notification + appearance preferences
    ├── submit/           Modal bottom sheet for sbatch job submission
    ├── navigation/       NavigationSuiteScaffold + NavGraph with shared-axis transitions
    └── theme/            SlrumManagerTheme (dynamic color, MotionScheme.expressive())

Stack summary:

Layer	Technology
UI	Jetpack Compose + Material 3 Expressive
Navigation	Navigation Compose 2.9 with NavigationSuiteScaffold
State	ViewModel + StateFlow
DI	Hilt 2.56
Local DB	Room 2.7
Preferences	Jetpack DataStore
Secure storage	androidx-security-crypto (EncryptedSharedPreferences)
Background work	WorkManager 2.10
SSH	sshj 0.38 + BouncyCastle 1.78 (full, not Android-stripped)
Serialization	kotlinx.serialization 1.8

---

Requirements

Android	11+ (API 30)
Target SDK	36
Cluster	Any host running Slurm with an SSH daemon
Auth	Password or private key (Ed25519 / RSA / ECDSA)

---

Building

1. Clone the repo:

   git clone https://github.com/Tresillo2017/slurm-manager.git
   cd slurm-manager

2. Open in Android Studio Meerkat (2025.1) or later.

3. Build and run on a device or emulator (API 30+):

   ./gradlew :app:assembleDebug

No API keys or local secrets are required — the app connects exclusively to servers you add.

---

Adding a server

1. On first launch, the onboarding wizard starts automatically.
2. Enter hostname, port (default 22), and username.
3. Choose Password or SSH Key authentication.
   • For SSH Key: tap Import Private Key and select your .pem / private key file.
4. Tap Test Connection — the app runs a live SSH handshake.
5. Tap Save Server on success.

Additional servers can be added at any time from the Servers screen → + button.

---

Polling & alerts

Polling is done via a WorkManager periodic task scheduled per server. The interval is configurable per server (1–60 minutes) from the server detail screen.

Alert rules are also per-server and support:

Alert type	Trigger
QUEUE_WAIT_EXCEEDED	Job pending longer than threshold
RUNTIME_EXCEEDED	Job running longer than threshold
JOB_FAILED	Job transitions to FAILED / CANCELLED
NODE_FAILURE	Node reports down in sinfo
PARTITION_DOWN	Partition reports down in sinfo

---

Contributing

Pull requests are welcome. For large changes, open an issue first to discuss what you'd like to change.

# Run unit tests
./gradlew :app:test

# Run instrumented tests (requires a connected device/emulator)
./gradlew :app:connectedAndroidTest

---

License

MIT © Tomás Palma