Hardware and system requirements¶
To ensure optimal performance for your monitoring environment, check the appropriate hardware specifications before installing PMM.
For guidance on selecting the best deployment method based on these requirements, see the choose your PMM deployment strategy topic.
PMM Server resource requirements¶
Resource requirements scale with the number of nodes and services monitored. Here are our recommendations for different deployment scales:
This is the most common deployment scenario, suitable for small to medium-sized environments:
- CPU: 4 cores
- Memory: 8 GB
- Storage: 100 GB
- Example workloads: Development environments, small businesses, initial deployments
Low-memory environments
With 8 GB RAM, you may need to switch ClickHouse to its low-memory configuration if you see “memory limit exceeded” errors. See ClickHouse memory issues.
Recommended for environments monitoring MySQL, PostgreSQL, or MongoDB at scale:
- CPU: 8-16 cores
- Memory: 16-32 GB
- Storage: 200 GB
- CPU usage: Expect 20-70% utilization
- Example workloads: Production environments, mid-sized companies
Designed for extensive monitoring environments with high-node counts:
- CPU: 16+ cores
- Memory: 32+ GB
- Storage: 500+ GB
- Example workloads: Large enterprises, mission-critical database fleets
Optimize ClickHouse for limited memory¶
If you experience ClickHouse memory issues (such as “memory limit exceeded” errors), you can switch to a low-memory configuration optimized for constrained environments. See ClickHouse memory issues for details.
Storage planning¶
Adjust storage calculations based on your data retention period and the number of metrics collected. To estimate storage requirements:
- Base formula:
nodes × retention_period_in_weeks × 1 GB - Quick estimate: for the default 30-day retention period, use the formula
number_of_nodes x 4 GB - High-precision monitoring: increase estimates by 20-50% when using 1-second collection intervals
Storage optimization¶
To reduce storage usage, consider disable table statistics, which can significantly decrease the size of the VictoriaMetrics database.
Architecture requirements¶
PMM Server¶
- CPU: must support the
SSE4.2, which is required for Query Analytics (QAN). - ARM64: starting with PMM 3.10.0, the PMM Server container image includes a native ARM64 build for ARMv8 or later. Docker automatically pulls the build that matches your host, so you don’t need to run the x86_64 image through emulation with
--platform linux/amd64. The PMM Server AMI is available for x86_64 only.
PMM Client¶
- Installation storage: Requires 100 MB of storage for installation
- Cache storage:
- VM Agent reserves 1 GB of disk space for caching during network outages
- Query Analytics (QAN) uses RAM instead of disk storage for its cache
- Architecture support: Compatible with both x86_64 and ARM64 architectures
- Operating systems: Compatible with modern 64-bit Linux distributions including Debian, Ubuntu, Oracle Linux, and “Red Hat” derivatives
For specific version support details, see Percona software support life cycle.
ARM-specific considerations¶
- Docker images: If using Docker for PMM Client on ARM systems, ensure you’re using the ARM64-compatible Docker images.
- Performance testing: Performance may vary across different ARM implementations. Conduct thorough testing to ensure optimal performance in your environment.
- Software compatibility: Ensure you’re using ARM-compatible versions of any additional software or databases you’re monitoring with PMM.
- Resource monitoring: Monitor resource usage closely on ARM systems, as it may differ from x86_64 systems. Adjust your configuration as needed for optimal performance.