telegraf
CLI Toolinfluxdata/telegraf
Agent for collecting, processing, and writing metrics, logs, and data.
Overview
Telegraf is a plugin-driven server agent for collecting and sending metrics and events from databases, systems, and IoT sensors. It supports over 300 plugins for inputs, outputs, and aggregators, compiles to a static binary with no external dependencies.
README Preview
# Telegraf\n\n[](https://godoc.org/github.com/influxdata/telegraf)\n[](https://hub.docker.com/_/telegraf/)\n[](https://goreportcard.com/report/github.com/influxdata/telegraf)\n[](https://circleci.com/gh/influxdata/telegraf)\n\nTelegraf is an agent for collecting, processing, aggregating, and writing\nmetrics, logs, and other arbitrary data.\n\n* Offers a comprehensive suite of over 300 plugins, covering a wide range of\n functionalities including system monitoring, cloud services, and message\n passing\n* Enables the integration of user-defined code to collect, transform, and\n transmit data efficiently\n* Compiles into a standalone static binary without any external dependencies,\n ensuring a streamlined deployment process\n* Utilizes TOML for configuration, providing a user-friendly and unambiguous\n setup experience\n* Developed with contributions from a diverse community of over 1,200\n contributors\n\nUsers can choose plugins from a wide range of topics, including but not limited\nto:\n\n* Devices: [OPC UA][], [Modbus][]\n* Logs: [File][], [Tail][], [Directory Monitor][]\n* Messaging: [AMQP][], [Kafka][], [MQTT][]\n* Monitoring: [OpenTelemetry][], [Prometheus][]\n* Networking: [Cisco TelemetryMDT][], [gNMI][]\n* System monitoring: [CPU][], [Memory][], [Disk][], [Network][], [SMART][],\n [Docker][], [Nvidia SMI][], etc.\n* Universal: [Exec][], [HTTP][], [HTTP Listener][], [SNMP][], [SQL][]\n* Windows: [Event Log][], [Management Instrumentation][],\n [Performance Counters][]\n\n## 🔨 Installation\n\nFor binary builds, Docker images, RPM & DEB packages, and other builds of\nTelegraf, please see the [install guide](/docs/INSTALL_GUIDE.md).\n\nSee the [releases documentation](/docs/RELEASES.md) for details on versioning\nand when releases are made.\n\n## 💻 Usage\n\nUsers define a TOML configuration with the plugins and settings they wish to\nuse, then pass that configuration to Telegraf. The Telegraf agent then\ncollects data from inputs at each interval a
FAQ (5)
TroubleshootingHow to fix Telegraf opcua_listener permanent disconnect with StatusBadNodeIDUnknown error?
Upgrade to Telegraf v1.40.0 or later. The bug is in gopcua's reconnect state machine: when a subscription fails to recreate during auto-reconnect (e.g., StatusBadNodeIDUnknown 0x80340000 because a monitored node no longer exists in the server address space), the client is left in Connected state with no active subscriptions and a dead publish loop, so Telegraf never retries and stays disconnected until restarted. The fix (gopcua/opcua#877) landed in gopcua v0.9.1, which ships with Telegraf v1.40.0+. Until you can upgrade, the only workaround is restarting the Telegraf service when the disconnect occurs (e.g., via a watchdog/monitoring alert on the no subscriptions to resume debug log).
TroubleshootingHow to fix Telegraf OPC UA listener 'EOF' error when monitoring over 1150 nodes?
This is a known bug (issue #14568) where the OPC UA listener plugin sends too large a request when many nodes are configured with a secure connection. The server rejects it with 'BadTcpMessageTooLarge', causing an 'EOF' error. Workaround: split your nodes into groups of around 500 across multiple [[inputs.opcua_listener]] instances or group blocks. A permanent batching fix is being developed.
TroubleshootingHow do I fix the Telegraf nightly Docker image showing an old or stale version?
This was caused by a stale public distribution/origin artifact after the nightly CI and S3 publish had completed successfully; it was not a Docker build failure. InfluxData resolved the distribution issue. Update the image with:
docker run --pull=always --rm quay.io/influxdb/telegraf-nightly:alpine --versionIf the old image is still used, remove it and pull it again:
docker image rm quay.io/influxdb/telegraf-nightly:alpine
docker pull quay.io/influxdb/telegraf-nightly:alpineDuring the incident, the affected image reported Telegraf 1.39.0-f6fa1073 instead of the expected nightly commit 511c852e. As a temporary workaround, use the artifact from the latest successful CircleCI nightly job or build Telegraf from the required commit.
TroubleshootingWhy is Telegraf's inputs.nftables plugin very slow when reading large nftables sets?
The plugin was running nft without --terse, causing it to dump every element in every nftables set. On large or contended sets this can take minutes or hours. Upgrade Telegraf to a release that includes the inputs.nftables --terse fix. The plugin then passes --terse when "sets" is not included in include, for example:
[[inputs.nftables]]
include = ["tables", "chains"]If you need set metrics, upgrade nft to 1.1.7 or newer. That version can return set element counts with --terse, avoiding a full element dump. Verify with nft --version. On older Telegraf or nft releases, the temporary workaround is to avoid monitoring sets or reduce the size of the sets being inspected.
TroubleshootingWhy is the Telegraf system uptime metric reported as a gauge type in Prometheus?
A bug in Telegraf (unreleased 1.39.0) incorrectly registered the 'uptime' metric as a gauge. The correct type is counter since uptime only increases. This issue will be fixed before the release. For affected development builds, a workaround is to override the metric type in Prometheus using metric_relabel_configs or relabeling rules to treat it as a counter. Upgrade to the latest stable version once fixed.