Consult the reference for the metrics reported by the System Metrics Source on Linux.

Events generated by the System Metrics Source have metrics metadata to designate dimension and metric fields. The host field contains the hostname, and is included as a dimension in all of them. The collectors include:

• System
• CPU
• Memory
• Network
• Disk
• Container
• Process
• GPU

In the Source’s configuration modal, you can set the level of detail for each type of metrics:

• Basic enables minimal metrics, averaged or aggregated.
• All enables full, detailed metrics, specified for individual CPUs, interfaces, and so on.
• Custom displays sub-menus and buttons from which you can choose a level of detail (Basic, All, Custom, or Disabled) for each type of event.
• Disabled means that no metrics will be generated.

Basic and Custom have different meanings depending on event type and will be described under each section below.

The tables outline the metrics emitted for each mode (Basic or Custom) and where applicable, the dimensions (to indicate where the metrics are coming from).

System ​

With System Metrics enabled, Cribl Edge captures CPU load averages, uptime, and count. The Custom option allows you to include process metrics that reflect the numbers of processes in various states.

Metrics for the overall system include the following:

CPU ​

Cribl Edge captures active, user, system, idle, and iowait percentages over all CPUs, with options to add per-CPU metrics and raw time counters for each state.

Metrics for CPUs include the following:

• The Per CPU option adds metrics with the cpu dimension.
• The Detailed metrics option adds metrics with mode dimension set to: irq, softirq, steal, guest, guest_nice, and nice.

Memory ​

With System Metrics enabled, Cribl Edge captures memory metrics including total, used, available, swap_free, and swap_total, with the option to toggle all memory states.

Metrics for memory include the following:

Network ​

With System Metrics enabled, Cribl Edge captures bytes, packets, errors, and network connections over all interfaces. The Custom option allows you to filter interfaces, and to decide whether to select per-interface metrics and generate protocol metrics.

Metrics for networks include the following:

Disk ​

With System Metrics enabled, Cribl Edge captures disk-used metrics - in percent, bytes read and written, and read and write operations - over all mounted disks. The Custom option allows you to filter devices, mountpoint, and filesystem type, and to decide whether to select per-device metrics and generate detailed metrics.

Metrics for Disk include the following:

Container ​

With System Metrics enabled, Cribl Edge generates Docker information with CPU, memory, network, and disk metrics for running containers. Optionally, you can customize which containers to generate metrics from.

Metrics for Container include the following:

Process Metrics ​

With Process Metrics enabled, Cribl Edge captures process-specific metrics from Linux servers and reports them as events. This allows you to monitor specific processes on Cribl.Cloud instances. You can generate events for any process object.

For information on how to configure the System Metrics Source to generate process-specific metrics, check out the Process Metrics section of the System Metrics page.

Process-specific metrics are not affected by the Host Metrics detail setting.

Process-specific metrics include the following:

GPU Metrics ​

With GPU Metrics enabled, Cribl Edge captures temperature, utilization, memory, power, clock, and throttle metrics from Nvidia GPUs.

The All option adds per-GPU events with identifying dimensions (gpu_index, gpu_name, gpu_uuid) and detailed PCIe, ECC, encoder, and power metrics.

In Basic mode, or when Per GPU metrics is disabled in Custom mode, the collector emits a single aggregated metric event for all GPUs during each collection interval. These metrics use the same name with the _all suffix, such as node_gpu_temperature_celsius_all. Across GPUs, those _all values use sum, average, or maximum, depending on how the metric is defined. The node_gpu_count gauge appears only on the aggregated event and does not have an _all equivalent.