Objective

This article provides guidance on how to estimate the resource requirements (CPU, memory, and disk) for the YugabyteDB Anywhere (YBA) node, specifically for the Prometheus monitoring service it runs.

Applies To

• Product: YugabyteDB Anywhere
• Version: All versions

Procedure

Accurately sizing your YugabyteDB Anywhere node is critical for ensuring a stable and performant monitoring system. The primary driver of resource consumption is the total number of metrics collected from all the YugabyteDB universes you manage.

Option 1: Use the Interactive Sizing Calculator (Recommended)

For the fastest and most straightforward estimation, use the interactive online calculator.

Launch the Prometheus Sizing Calculator: https://yugabyte.github.io/yugabyte-prometheus-sizing-calculator/

GitHub Repo: https://github.com/yugabyte/yugabyte-prometheus-sizing-calculator

How to Use the Calculator

1. For each YugabyteDB universe you plan to monitor, enter its Name, total Number of Tables, and total Number of Nodes (T-Servers).
2. Click Add Universe.
3. Repeat for all universes managed by this YBA instance. The Estimated Requirements will update automatically. See following demo (available on YouTube as well). 

Option 2: Use the Script-Based Calculator

If you prefer to use a script, Yugabyte Support Team also provides a command-line calculator. You can just download the attached shell script or latest GitHub release and run on your local machine. 

prometheus-sizing-calculator.zip

Example Command

Usage: prometheus-sizing-calculator.sh -u 'name1:tables1:nodes1' -u 'name2:tables2:nodes2' ...

bash prometheus-sizing-calculator.sh -u 'Yugabyte-Dev-Cluster:1000:3' -u 'prod:500:6'

Sample Output

Processing Universes...
  - Universe: Yugabyte-Dev-Cluster - Tables: 1000, Nodes: 3 → Metrics: 207000
  - Universe: prod - Tables: 500, Nodes: 6 → Metrics: 234000

Total Tables: 1500 | Total Nodes: 9
Total Metrics: 441000

* Estimated Requirements:

  Memory: 12.61 GB
  Disk:   106.40 GB
  CPU:    .44 cores

Option 3: Manual Calculation

If you prefer to calculate the requirements manually, follow the steps below. These formulas assume a standard Prometheus configuration with a 10-second scrape interval and a 15-day data retention period.

Step 1: Calculate Total Metrics

First, calculate the metrics for each universe using the formula below, then add them all together to get your Total Metrics:

Universe Metrics = (60 * Number of Tables + 9000) * Number of Nodes

Step 2: Estimate Required Resources

Use your Total Metrics value from Step 1 to calculate the resource estimates.

• Memory (GB):

  • If Total Metrics > 1,000,000:
    Memory (GB) = (Total Metrics * 10) / 1024 / 1024
  • If Total Metrics ≤ 1,000,000:
    Memory (GB) = (Total Metrics * 30) / 1024 / 1024

• Disk (GB):
  Disk (GB) = (Total Metrics * 253) / 1024 / 1024

• CPU (Cores):
  CPU Cores = Total Metrics / 1,000,000

Important Considerations and Final Recommendation

The calculated values are a baseline for the Prometheus service alone.
Actual resource usage on the YugabyteDB Anywhere node will be higher due to several dynamic factors.

Understanding Memory Consumption

Estimating Prometheus’s memory needs is complex. The following factors can significantly increase consumption:

• Query Load: High PromQL query volumes from multiple users accessing dashboards.
• System State: Temporary memory spikes during restarts or when new tables are loaded quickly.
• Metrics Volume: YugabyteDB’s distributed architecture exposes a vast number of metrics.
• Federation: Using Prometheus federation for High Availability can further amplify memory demands.

Final Recommendation

In addition to Prometheus, the YugabyteDB Anywhere host also runs:

• The core YBA Platform service
• A PostgreSQL database for metadata

These services consume their own resources.

Taking all these factors into account, we strongly recommend allocating at least 30% additional memory to Prometheus calculation for the YugabyteDB Anywhere host. This provides a safe buffer (30–40%) over the calculated Prometheus baseline to accommodate the full YBA software stack and ensure system stability.