Instructor-led training / klique-sa
Klique Systems Administrator
klique-sa
- Duration
- 3 Days
- Format
- Live Online
- Access
- Multiple Timezones
Overview
About this course.
The Klique Systems Administrator course at Alta3 Research Labs is a definitive, hands-on training program for Systems Administrators and DevOps Engineers tasked with deploying and securing enterprise-grade MLOps environments. Moving beyond basic usage, students dive deep into the Alta3 lab environment—assisted by VIRGIL, our AI Lab Coach—to master the underlying topology of Klique (previously know as ClearML), including the complex relationships between Servers, Agents, and Kubernetes. You will learn to architect robust data flows, manage external data stores like MongoDB and Elasticsearch, and handle secrets securely, ensuring you understand not just how to use the platform, but how to own the infrastructure that powers it.
A critical focus is placed on transforming standard deployments into secure, multi-tenant ecosystems suitable for large organizations. Participants will master identity integration via OIDC and SSO, implement strict Role-Based Access Control (RBAC), and govern isolated tenants mapped to specific resources. The training culminates in a rigorous Capstone Lab covering "Day-2" operations, where you will deploy Prometheus and Grafana for observability, troubleshoot agent queues, and provision a new tenant end-to-end. By the end of this course, you will possess the specialized skills to architect and maintain a resilient, scalable Klique platform that satisfies enterprise security and operational standards.
Review this course online at https://www.alta3.com/courses/clearml-sa
Objectives
What you'll learn
- Klique Architecture Topology
- Secure Multi-Tenancy Implementation
- Identity & Access Governance
- Kubernetes & Agent Orchestration
- Observability & System Monitoring
- Enterprise Operational Troubleshooting
Audience
Who it's for
- System Administrators
Prerequisites
Before you start
None
Outline
Course outline.
Welcome to Klique Systems Administrator
- 💻 Exploring Your Lab Environment
- 💻 Meet VIRGIL: Your AI Lab Coach
- 💻 Register for Polls
Core Klique Platform Operations
- 💬 The Role of Klique in the MLOps Ecosystem
- 💬 Kubernetes Requirements and Ingress Setup
- 💻 Install Kubernetes using Ansible
- 💻 Persistent Storage for Klique
- 💻 Prepare Ingress Controller for Klique
- 💬 Klique Helm Chart Overview and Customization
- 💻 Deploy Klique via Helm
- 💬 Connecting External Stores and Managing Secrets
Use Klique for Tracking & Orchestration
- 💬 Klique Architecture: Clients, Server, Agents
- 💬 Tasks, Projects, and Artifacts Explained
- 💻 Klique Quickstart: Log a Training Run
- 💻 Explore Tasks, Projects, and Artifacts in Klique
Understand Klique Topology (Stores, Queues, Security)
- 💬 Klique Core Components and Data Flow
- 💬 External Data Stores: MongoDB, Elasticsearch, Redis
- 💻 Explore Klique Architecture Topology
- 💬 Queues, Agents, and Cache Management
- 💻 Klique Agent: Kubernetes Glue
- 💬 Security Foundations: TLS, OIDC, RBAC
- 💻 Enable TLS and Validate Secure Endpoints
Deploy Tenant Services (Multi-Tenancy & Web Authentication)
- 💬 Understanding Multi-Tenancy in Klique
- 💬 Web Login and Authentication Options (SSO / OIDC / SAML)
- 💻 Enable Multi-Tenant Mode via Helm
- 💻 Create and Manage Tenants via Klique SDK
- 💻 Map Tenants to Kubernetes Namespaces & Apply Policy
Identity & Authentication (OIDC / SSO / SAML)
- 💬 Identity Provider Integration: OIDC, SAML, and Authentication Flows
- 💬 Klique Authentication Modes: Internal, External, Hybrid
- 💻 Deploy and Initialize KeyCloak
- 💻 Configure Klique with OIDC (Keycloak / Azure AD Example)
- 💬 Troubleshooting Authentication & Login Failures
Tenant Administration & Governance
- 💬 Tenant Governance: RBAC, Groups, Permissions, Separation of Data
- 💻 User & Group Lifecycle: Provisioning, Access, Deactivation
- 💻 Inspect Tenant Activity, Audit Logs & Usage Patterns
- 💬 Tenant Lifecycle Best Practices (Provision → Operation → Decommission)
Queue & Agent Operations for Tenants
- 💬 How Klique Agents Work
- 💬 Queue Design: Shared Queues, Tenant Queues, GPU Queues
- 💻 Create/Modify Queues for Tenants
Dynamic GPU Fractions - CDMO
- 💬 Concepts & Mental Models (MIG, MPS, Fractions, Profiles)
- 💬 Architecture – GPU Operator, Device Plugin, and CDMO Integration
- 💬 Installation and Version Pinning
- 💻 Prepare GPU Nodes for the Klique Dynamic MIG Operator (CDMO)
- 💻 Deploy the NVIDIA GPU Operator
- 💬 Scheduling, Validation, and Observability
- 💻 Enable CDMO for MIG-Based Fractional GPU Scheduling
- 💬 Troubleshooting and Recovery
Klique Pipelines
- 💻 Initializing Pipelines with the Python SDK
- 💻 Pipelines: Build a Simple ML Workflow
- 💻 Pipelines: Run Tasks in Parallel
- 💻 Pipelines: Reuse Results with Caching
- 💻 Pipelines: Execute Remote Workloads from GitHub
Continuing Education
- 💻 Continuing Education (Klique in Practice)
- 💬 Curriculum Path: FILL OUT LATER
For teams
Training a whole team?
Request private training for Klique Systems Administrator and we'll follow up with a quote for your group.