dstat

dstat
Name	dstat
Author	H. Peter Anvin
Developer	Corentin Chary; community contributors
Released	2006
Latest release	0.7.3 (example)
Programming language	Python (programming language)
Operating system	Linux
Genre	System monitor
License	GNU General Public License

dstat

dstat is a versatile, real-time performance monitoring tool for Linux systems designed to replace and combine functionality found in vmstat, iostat, netstat, and ifstat. It provides per-interval reports of CPU, disk, network, paging, and process statistics and is commonly used by system administrators, performance engineers, and developers during capacity planning, troubleshooting, and benchmarking on servers such as those deployed by Amazon Web Services, Google Cloud Platform, and Microsoft Azure. dstat's extensible plugin architecture and scriptable output make it interoperable with tools like Collectd, Graphite, Prometheus, and Grafana for long-term visualization and alerting.

Overview

dstat was initially developed by H. Peter Anvin in the mid-2000s to offer a unified replacement for multiple UNIX performance utilities including vmstat and iostat. The project attracted maintenance and feature contributions from the open-source community, including contributors affiliated with organizations such as Canonical (company) and commercial users at Facebook, Twitter, and Netflix who integrate dstat into performance workflows. The tool runs on Linux kernels and interacts with kernel interfaces exposed by components such as procfs and sysfs to sample metrics. dstat is distributed under the GNU General Public License and is available in package repositories for distributions including Debian, Ubuntu, Red Hat Enterprise Linux, and Fedora.

Features

dstat exposes multiple columns of live metrics: CPU utilization, disk I/O, network throughput, page faults, and system interrupts. It supports plugin extensions which allow gathering metrics from sources including MySQL, PostgreSQL, MongoDB, Redis (database), Docker (software) containers, and virtualization layers like KVM and Xen. Output formats include human-readable tables, CSV for ingestion by Excel, LibreOffice, or data pipelines feeding InfluxDB and Graphite. dstat provides options for per-process statistics traceable to processes started under systemd, Upstart, or legacy init systems; it can correlate hotspots with transaction traces produced by application profilers such as perf (Linux tool). The plugin API is written in Python (programming language), enabling integration with automation tools like Ansible, Puppet (software), and Chef (software).

Usage

dstat is invoked from a shell on Linux distributions to collect live statistics over specified intervals and counts. Common invocation patterns combine flags to select CPU, disk, and network metrics simultaneously, and users often pipe dstat's CSV output into tools like rsyslog, Fluentd, or Logstash for aggregation. Administrators use dstat during load-testing with benchmarking tools such as ApacheBench, wrk (software), sysbench, and JMeter to correlate system-level metrics with application throughput. For remote monitoring, dstat is run over secure connections provided by OpenSSH and integrated into continuous integration systems like Jenkins and GitLab CI for regression detection of performance anomalies.

Examples

- Simple live monitoring: launch dstat to display CPU, disk, and network columns during a workload generated by Apache HTTP Server or Nginx and observe correlations between request rate and I/O wait. - CSV logging: run dstat with the --output option and import results into LibreOffice or feed into Graphite using a custom script to create dashboards in Grafana that compare metrics across instances in a Kubernetes cluster. - Plugin use: enable a MySQL plugin to monitor MySQL counters (connections, queries per second) while running a sysbench OLTP benchmark; use the data to tune buffer sizes and scheduler parameters exposed by Linux kernel tunables. - Containerized environments: execute dstat inside Docker (software) containers or on hosts orchestrated by Kubernetes to distinguish host-level NIC contention from container-level CPU throttling.

Implementation and Development

dstat is implemented in Python (programming language) and interacts with kernel interfaces like procfs, sysfs, and netlink to sample metrics. The plugin system exposes a Python API for adding collectors that read from sources such as /proc/net/dev and /proc/diskstats or query external services using clients for MySQL, Redis (database), and HTTP endpoints. Development has occurred on public code hosting platforms similar to GitHub and has attracted contributions from engineers at companies including Canonical (company), Red Hat, and cloud providers. Maintenance efforts typically address compatibility with evolving Linux kernel releases, Python 2/3 interoperability, and packaging across distributions like Debian and Fedora.

Comparison with similar tools

dstat combines features found in utilities such as vmstat, iostat, netstat, and ifstat into a single consistent interface, distinguishing it from single-purpose tools. Compared with daemon-based collectors like Collectd and Telegraf, dstat is designed for interactive, ad-hoc investigation rather than continuous background collection; it complements observability stacks built around Prometheus and Graphite by providing immediate on-host diagnostics. For deeper profiling, tools like perf (Linux tool) and SystemTap provide event-level sampling and tracing, whereas dstat focuses on aggregated counters and throughput metrics useful during benchmarking with ApacheBench or sysbench.

Category:Linux software