Centralized Performance Monitoring
Cockpit provides performance monitoring and metrics visualization across physical host servers and virtual machines. The monitoring subsystem aggregates CPU, memory, storage, and network statistics to assist with capacity planning and performance optimization.
Metrics Collection and Downsampling
The metric collection engine manages data aggregation and storage efficiency:
- Metrics Collection: The collector service queries physical hosts and virtual machines at a default 10-second polling interval.
- In-Memory Buffering: Collected data is temporarily buffered in memory to optimize write operations to the database.
- Downsampling Policy: To manage storage utilization, Cockpit implements a data downsampling policy. Raw 10-second samples are stored for a short duration (e.g., 24 hours). Historical data is subsequently aggregated into hourly and daily averages, allowing long-term trend analysis without degrading database read performance.
Dashboard Implementations
Performance data is visualized across three management scopes:
1. Cluster Overview Dashboard
Presents aggregated resource metrics for the entire virtualization environment:
- Total Compute Allocation: Displays total CPU cores and RAM capacity across all active hypervisors, alongside current utilization percentages.
- Aggregated Storage: Lists total capacity, used blocks, and remaining storage space across all datastores.
- Host Node Registry: A summary showing the operational connection status (
connected,disconnected,error) for all physical nodes.
2. Host Detail Dashboard
Displays performance telemetry for a specific physical host:
- Real-Time Visualization: Continuous line charts showing processor utilization, memory allocations, and disk I/O throughput.
- System Profile: Identifies host hardware configurations (such as processor type, BIOS version, OS kernel release, and system uptime).
3. Virtual Machine Telemetry
- VM List Sparklines: Inline trend indicators displayed next to VM names in lists. These provide immediate visual context on workload patterns without requiring you to open the VM's detail page.
- VM Metrics: Detailed performance histories, virtual disk bandwidth, and network packet rates for a single virtual machine instance.