Skip to content

One collector for every signal, run by policy

The Tracix Collector is an OpenTelemetry Collector distribution for Linux, macOS, Windows, Docker and Kubernetes. Each collector enrolls with a policy that decides what it collects. Configuration arrives signed, over a connection the collector opens itself, and upgrades roll out from the console.

  • Linux · macOS · Windows · Docker · Kubernetes
  • amd64 and arm64
  • Ed25519-signed configuration

How it works

From one command to a managed fleet

Every connection starts at the collector, so nothing on your network has to accept a connection from Tracix.

  1. 01InstallOne command on a host, or the Helm chart in a cluster, with every release checked against its SHA-256.
  2. 02EnrollA policy's token enrolls the collector, which gets its own identity and deletes the token file.
  3. 03ConfigureThe policy's integrations and outputs render into one configuration, signed with Ed25519.
  4. 04DeliverA WebSocket push triggers a fetch of the new revision; polling every 60 seconds is the fallback.
  5. 05ApplyEach configuration is validated before it goes live, and a failure falls back to the last good one.
  6. 06UpgradeUpgrade one collector or 100 at once, or let the policy keep its collectors on the stable channel.

Install

Runs where your workloads already run

Install with one command on Linux, macOS or Windows, run the container image, or deploy the Helm chart as a DaemonSet. The collector needs outbound HTTPS on port 443, directly or through your proxy, and the installer tests that path first, stopping with the cause when DNS, a firewall or a TLS-inspecting proxy is in the way.

  • Linux, macOS and Windows on amd64 and arm64, plus container images for both
  • On Linux it runs as an unprivileged user, and by default its only extra capability lets it receive SNMP traps
  • Air-gapped hosts can install from your own mirror, pinned to a SHA-256 checksum
$ curl -fsSL https://charts.tracix.io/install.sh | sudo bash -s -- \  --endpoint https://agents-acme-prod-x7k2.app.tracix.io \  --token-file /root/tracix-token==> Tracix Collector — Linux install==> verified SHA-256 checksum==> starting tracix-collector.service==> enrolled; waiting for the policy configuration==> Installation complete!
Policy linux-servers · r14 · signed128 collectors running

Policies

Configuration that is signed, checked and acknowledged

Every collector belongs to one policy, chosen by the token it enrolled with. Saving the policy creates a new revision. Collectors fetch it, verify its Ed25519 signature against the key they pinned at enrollment, validate it, and report back only once their pipelines are running.

  • Revisions

    Each save is a numbered revision, and two people editing at once cannot overwrite each other.

  • Signed

    Collectors reject any configuration that does not carry the workspace's signature.

  • Push, then poll

    WebSocket push for speed, with polling every 60 seconds when the socket is down.

  • Safe apply

    A receiver that fails to start is switched off and reported; other failures restore the last good config.

Upgrades

Upgrade the whole fleet from one screen

Upgrade a single collector, up to 100 at once, or let the policy follow the stable channel. Hosts download the new release, verify its checksum, validate the binary and swap it in, keeping the previous one. Kubernetes collectors roll their own DaemonSet one node at a time.

  • On Windows the previous binary comes back if the service does not start
  • Docker collectors recreate themselves on the new image when the Docker socket is mounted
  • Downgrades are refused, and with auto-update on, a failed update is retried after 15 minutes, an hour and four hours

Supported platforms

Where the collector runs

Every platform gets the same distribution and the same policy model. A few integrations need a specific operating system.

PlatformArchitecturesInstall withRuns as
Linuxamd64, arm64install.shsystemd service, unprivileged user
macOS 12 and laterIntel, Apple siliconinstall.shlaunchd daemon
Windows Server 2016 and Windows 10 or lateramd64, arm64install.ps1Windows service
Dockerlinux/amd64, linux/arm64docker run with a digest-pinned imageNon-root container
KubernetesLinux nodes, amd64 and arm64Helm chartDaemonSet, read-only root filesystem

Linux means Ubuntu 20.04+, Debian 11+, RHEL and CentOS 8+ or Amazon Linux 2 with systemd. Zero-code eBPF tracing, eBPF profiling, host security, software inventory and journald need Linux; Windows Event Log needs Windows. There are no deb, rpm or MSI packages: the installers download the release archive for your platform and verify it.

FAQ

Questions platform teams ask first

Is the Tracix Collector OpenTelemetry?
Yes. It is built with the OpenTelemetry Collector Builder from upstream core and contrib components, plus Tracix extensions for fleet management, discovery and network jobs. It speaks standard OTLP, so your SDKs and existing collectors can send to it.
What does the collector need from my network?
Outbound HTTPS on port 443 to your workspace and the release host, directly or through a proxy. Tracix never opens a connection to the collector; listeners such as syslog or SNMP traps are only for devices on your network.
Can I restart a collector from the console?
Not yet. Today the console upgrades, unenrolls and uninstalls collectors, one at a time or up to 100 at once. Remote restart is on the roadmap.
Can I roll back a policy from the console?
No. Every save is kept as a numbered revision, and each collector keeps its last five good configurations on disk, but the console has no button to restore an earlier revision today.

Install your first collector in minutes

Create an enrollment token for a policy, run one command, and the collector appears in the fleet with its policy and health.