Fleet Management

Fleet Management lets you manage fleets of OpenTelemetry collectors from one console over the OpAMP protocol. Enroll collectors with tokens, track their health and version, distribute versioned collector configurations, and roll changes out in phases with health gates and automatic rollback.

Enable

Fleet Management is an Enterprise feature and is on by default. Set O2_FLEET_ENABLED=false to turn it off; reads still work, writes are refused, and the OpAMP endpoint answers 503. O2_OPAMP_ENABLED remains as an alias.

Enroll collectors

A collector joins through an enrollment token. The token's plaintext (o2ft_...) is shown only once, at creation — store it somewhere secure.

1. In Fleet Management, create a token.
2. Set its name, the fleet the collector joins, and optional expiry and max uses.
3. Copy the install command and run it on the collector host.

Revoke a token to stop it from enrolling new collectors.

Fleets

A fleet groups collectors that share a configuration. Every org starts with a default fleet; create more to stage changes or split environments. Move collectors between fleets to re-assign their configuration. Delete a fleet only after moving out its collectors and revoking its tokens.

Collectors

The Collectors list shows every enrolled collector with its status (online, offline, degraded, unknown), version (flagging anything behind the target version), host, OS/architecture, and fleet. Open a collector to see its capabilities, component health, effective configuration, delivery evidence, and event history. From the detail drawer you can rename, restart, redeliver its configuration, or delete it.

Configurations

Configurations are versioned YAML collector configs. Create a config, edit the YAML, and validate it before publishing; a config with validation errors cannot be published. Publishing a version makes it available to fleets. The restricted components list in settings blocks component types you do not want used.

A visual editor shows the pipeline as nodes — receivers, processors, and exporters — and lets you add components and pipelines without editing YAML directly.

Rollouts

A rollout assigns a configuration version to a fleet. Choose whether it starts immediately or at a scheduled time, and a batching plan:

• Canary — a small first phase, then the remainder.
• Progressive — an initial batch that grows by a multiplier, capped at a maximum.

The rollout converges phase by phase, pausing at health gates; offline collectors are left out at the start. If a rollout shows evidence of harm it halts with a class:

• proven_bad — collectors rejected the config.
• looks_worse — the health gate failed.
• cannot_see — no data, a timeout, or collectors offline.

Roll a halted rollout back manually, or let it auto-rollback, which is pre-authorized for proven_bad and looks_worse.

Settings

Set the org target collector version — collectors on an older version are flagged outdated. Per fleet, tune the heartbeat interval, offline threshold, offline ratio, and phase timeout, which drive when a collector is considered offline and how rollouts proceed.

Audit trail

Every fleet action — tokens, configurations, rollouts, and collector actions — is written to the audit stream. Review it from the Audit tab.

Permissions

RBAC is enforced through fleets (per fleet) and fleet_configs / fleet_management (org level). Grant users only the fleets they should see; the collectors, configurations, and tokens under a fleet follow the fleet grant.