Prometheus
Prometheus is required when operating Bindplane in high availability.
Bindplane stores time series metrics in Prometheus. These time series metrics allow Bindplane to track collector throughput measurements over time. When viewing the summary page or a specific configuration, the measurement data being displayed is the work of the measurement database.
When operating Bindplane in single instance mode, a bundled version of Prometheus will be used. The user does not need to install or configure Prometheus.
Prerequisites
Sizing
Prometheus can be scaled vertically based on the number of managed collectors. The volume of time series metrics being pushed to Prometheus will scale linearly with the number of collectors.
The following table provides a general guideline for sizing your Prometheus instance based on the number of collectors. This table should service as a starting point for sizing your Prometheus instance.
| Collector Count | CPU Cores | Memory | Disk (GB) |
|---|---|---|---|
| 1-25,000 | 2 | 8GB | 80 |
| 25,000-250,000 | 4 | 16GB | 300 |
| 250,000-500,000 | 8 | 32GB | 600 |
| 500,000-1,000,000 | 16 | 64GB | 1200 |
note
Complicated configurations with many sources, processor nodes, and destinations will output more time series metrics.
Installation
Prometheus should be installed on a dedicated system. The installation will be accessed by each Bindplane server in your environment.
Follow the installation documentation for instructions for deploying a shared Prometheus instance.
Configuration
Bindplane will use a bundled version of Prometheus by default. The configuration must be updated to use a remote Prometheus instance. See the configuration configuration documentation for instructions on configuring your Bindplane servers to use the shared Prometheus instance.