Metrics Reports

You can use gadmin metric commands to collect the following metrics related to state of resource usage on your instances:

  • CPU usage

  • Memory usage

  • Disk space usage

  • Network-related information

For a full list of options available to gadmin metric, see gadmin metric or run gadmin metric -h in the command line to view the help text for the command.

TigerGraph also provides log files.

Report CPU usage

Use the command gadmin metric -t cpu to report CPU usage by different TigerGraph services on a cluster or a node in the cluster.

To report CPU usage on select nodes, use the --host flag or the -m shorthand, followed by the nodes for which you want CPU usage data, separated by commas. If you choose not to specify nodes, then the command reports CPU usage for all nodes in the cluster.

To select specific services to report data for, specify the service IDs after the command. If you don’t specify a service, the command reports data for all services.

Service ids currently include the following: gpe gse restpp gsql ifm gui ctrl kafka etcd zk nginx dict admin kafkaconn kafkastrm-ll

The following command reports CPU usage by the GUI service for m1 and m2:

$ gadmin metric gui -t cpu -m m1,m2

The following command reports CPU usage by all services for m1 and m2:

$ gadmin metric gui -t cpu -m m1,m2

Report memory usage

Use command gadmin metric -t mem to report memory usage on a cluster or a node in the cluster.

To report memory usage on select nodes, use the --host flag or the -m shorthand, followed by the nodes for which you want memory usage data. If you choose not to specify nodes, then the command reports memory usage for all nodes in the cluster.

To select specific services to report data for, specify the service IDs after the command. If you don’t specify a service, the command reports data for all services.

Service ids currently include the following: gpe gse restpp gsql ifm gui ctrl kafka etcd zk nginx dict admin kafkaconn kafkastrm-ll

The following command reports memory usage by the GUI service for m1 and m2:

$ gadmin metric gui -t mem -m m1,m2

The following command reports memory usage by all services for m1 and m2:

$ gadmin metric -t mem -m m1,m2

Memory usage reported by gadmin metric -t mem may appear higher when TigerGraph runs in containers than when it runs directly on a host. This difference is expected and is caused by how memory usage is measured in container environments.

Memory usage differences between container and host deployments

TigerGraph reports memory usage based on the underlying runtime environment. Container-based deployments and host-based deployments use different memory accounting models.

Deployment type Memory metric used How memory is counted

Container (Kubernetes)

Working set memory

Includes active page cache. Cached memory is counted as used memory.

Host (non-container)

Resident/used memory

Active page cache is typically excluded and treated as reclaimable.

Container (Kubernetes) deployments

  • Kubernetes monitoring tools typically report working set memory.

  • The working set includes active page cache.

  • For example, kubectl top pods reports working set memory.

  • As a result, memory usage values appear higher.

Host (non-container) deployments

  • Host-level tools such as free -g and top usually do not count active page cache as used memory.

  • Cached memory is excluded from “used” memory totals.

TigerGraph memory metrics behavior

TigerGraph follows the same approach as the underlying environment:

  • In container environments, gadmin metric -t mem reports working set memory, including active page cache.

  • In host environments, the reported memory usage excludes active page cache.

Because of this difference, memory usage reported by:

$ gadmin metric -t mem -m m1,m2,m3,m4

may look higher in container-based deployments than in host-based deployments, even when the actual workload and data volume are similar.

This behavior is expected and does not indicate higher real memory pressure.

Report disk space usage

Use command gadmin metric -t disk to report disk space usage for a cluster or select nodes in a cluster.

To report network information on select nodes, use the --host flag or the -m shorthand, followed by the nodes for which you want disk usage data. If you choose not to specify nodes, then the command reports data for all nodes in the cluster.

The following command reports disk usage for m1 and m2:

$ gadmin metric -t disk -m m1,m2

Report network connections and traffic

Use command gadmin metric -t net to report TCP connections and network I/O for a cluster or select nodes in a cluster.

The command reports three data points for each node:

  • TCP CONNS: number of open connections.

  • IN: Network inflow since the node started.

  • OUT: Network outflow since the node started.

To report network information on select nodes, use the --host flag or the -m shorthand, followed by the nodes for which you want network data. If you choose not to specify nodes, then the command reports data for all nodes in the cluster.

The following command reports network information for m1 and m2:

$ gadmin metric -t net -m m1,m2