Glances 4.5.0: System monitoring with NPU and ZFS support

Contents

The cross-platform system monitoring tool Glances is available in version 4.5.0.1. The new release of the open-source project brings support for Neural Processing Units (NPUs), extended NVMe metrics, and native monitoring for ZFS. Version 4.5.0 was withdrawn shortly after its release due to critical errors and replaced by 4.5.0.1.

Changes in detail

For the first time, Glances can now monitor the performance of Neural Processing Units, which are used in modern processors for AI and machine learning workloads. For NVMe drives, the software provides extended SMART metrics such as temperature, wear leveling information, and error statistics. Native monitoring of the ZFS file system includes the status of storage pools, scrub and repair operations, and I/O performance data.

In version 4.5.0, the developers fixed several security issues, including Jinja2 vulnerabilities that allowed potential server-side template injection. The Glances RestfulAPI and WebUI were affected. Further fixes address timeout issues and the logging of sensitive information in plain text. The security vulnerabilities were identified using the Bandit security tool. Users should update to version 4.5.0.1 immediately.

To secure the RestfulAPI server, Glances introduces token-based authentication. Configuration is done via the glances.conf file and protects against unauthorized access to the monitoring interface.

Bugfixes for WebUI and Exporter

Additionally, the developers have corrected numerous errors in the web interface. The CPU speed display now shows correct values. The TIME+ field in the process list no longer shows excessive values. Type incompatibilities for Active Monitoring Plugins (AMPs) have been resolved in the InfluxDB integration.

A new export option for DuckDB allows monitoring metrics to be written to the lightweight database. Configuration is done via a [duckdb] section in the configuration file. DuckDB is suitable for fast SQL-based ad-hoc analysis of time-series data without a separate database server. Persistent storage in DuckDB allows performance trends to be identified over longer periods – for example, to detect bottlenecks or patterns in system behavior.

Improved Hardware Support

The new release improves AMD GPU support through better detection of graphics card names via the amdgpu.ids file. In the process list, Glances now indicates which CPU core a process is running on. This information helps diagnose core bottlenecks, NUMA problems, and scheduling errors.

All information about Glances 4.5.0 or 4.5.0.1 can be found in the Release Notes on GitHub. Version 4.4.0 had already introduced a Python API and a snapshot mode.

Glances runs on GNU/Linux, BSD, macOS, and Windows. Installation is done via pip install --upgrade glances or as a Docker container via docker pull nicolargo/glances:latest. The Docker build pipeline has also been improved in the new version.

(fo)