System Metrics

Maintaining a healthy TigerGraph cluster requires consistent monitoring of resource usage. The gadmin metric command provides a unified interface for this data.

1. Monitoring CPU Usage

You can report CPU usage by specific service or across the entire cluster.

bashterminal
Copy
# Report CPU for all services on all nodes
gadmin metric -t cpu

# Report CPU specifically for the GPE service on node m1
gadmin metric gpe -t cpu -m m1

2. Memory Usage & Reporting

TigerGraph measures memory differently depending on the environment:

Environment	Metric Used	Behavior
Host (Bare Metal)	Resident Memory	Excludes active page cache.
Container (K8s)	Working Set	Includes active page cache (often looks higher).

Tracking Memory by Query

For deep performance tuning, you can track how much RAM a specific query consumes by searching the GPE logs:

bashterminal
Copy
grep -i "QueryMem" $(gadmin config get System.LogRoot)/gpe/log.INFO

3. Disk & Network I/O

Disk Usage

Check if your storage is approaching capacity:

bashterminal
Copy
gadmin metric -t disk

Network Traffic

Monitor TCP connections and data inflow/outflow:

bashterminal
Copy
gadmin metric -t net

The output includes:

• TCP CONNS: Number of active connections.
• IN/OUT: Total network traffic since the service started.

4. Critical Thresholds

By default, TigerGraph will abort all queries if system free memory drops below 10% (SysMinFreePct). This protects the system from crashing due to Out-Of-Memory (OOM) errors.

[!TIP] Use gadmin status -v to see real-time memory pressure and process IDs for every service.