Operations

What changes when an agent is managed

Which console actions your Manager takes over, what always stays local, the limits on what a Manager may configure, and how drift and start-up failures are handled.

An enrolled agent is managed while it holds a licence lease from your DPLens Manager. From then on its configuration, its secrets and its licence come from the Manager, and the console on the machine becomes a window onto the agent rather than a way to change it.

An agent that is enrolled but still pending approval or waiting for a seat is not managed yet — it runs under its local configuration exactly as before. See Enrolling an agent with DPLens Manager.

The console

The sidebar reads Managed by your Manager — local changes are disabled.

These actions are refused with "managed by …", and each refusal is recorded in the audit log under the account that tried:

Everything else works as before: every page, the live stream, testing a destination, previews, the audit log, and the Licence page, which shows the licence the Manager granted.

What always stays local

A few settings belong to the machine and are never taken from the Manager:

The Manager never sees them; the agent puts them back into every configuration the Manager sends.

What a Manager may and may not configure

Your Manager renders the configuration; the agent checks every one against its own rules before running it, and refuses — by name, with the position of the offending setting — anything outside them:

These install-time choices are never changed by the Manager. An administrator can change them on the machine with dplens.exe --enrol, but only while the agent is not enrolled — before its first enrolment, or after it has been detached.

Secrets delivered by the Manager

A secret the Manager sends is encrypted to this agent's own key and sealed to the machine like every other secret. It is also bound to the destinations it was sent for: the agent will use it only for those destination addresses, for the same kind of credential, and only while their TLS settings are unchanged. A configuration that tried to send the credential somewhere else, or to turn certificate verification off for it, is refused.

This protects your credentials if a Manager is ever compromised. It is not a barrier against an administrator on the machine, who can always read what the agent can read.

Nothing shows a secret's value — not the console, not the Manager.

Changes made on the machine anyway

Someone with administrator rights can still edit the configuration file on a managed machine. The agent does not undo the edit: it reports it to the Manager as drift at its next check-in, and putting it right is the Manager's decision — an ordinary re-apply of the configuration it intended.

A reboot or an upgrade never shows up as drift.

When a managed configuration cannot start

If, at start-up, the configuration the Manager last sent cannot run because a secret it depends on no longer matches it, the agent does not fail to start. It starts with collection held, reports the problem to the Manager, records it in the audit log, and waits. As soon as the Manager sends a corrected configuration, collection resumes. Pull-based sources lose nothing while it waits.

If the Manager is unreachable

Nothing stops. The agent keeps collecting and delivering under its last configuration, and keeps its counters and five-minute summaries — up to a day of them — to send when it reconnects. Its licence lease lasts its term plus a grace period (by default 30 days plus 7), and only after that does collection stop, exactly as when a standalone licence expires. See Licensing.

When an agent stops being managed