Skip to main content

Frequently asked questions

Does the agent ever restart or reload my services?​

No. The agent never restarts, reloads or reconfigures a service of yours, and the console has no button that would make it. Changes that need a reload (enabling APM for a PHP version, for example) are staged on the host and applied only when you run the prepared command yourself. The one thing the agent restarts is its own service, after a self-update.

Why does the agent run as root?​

Reading every service's configuration and logs, following the journal, sampling processes and attaching eBPF probes all need root. The service unit confines what root can do: the system tree is read-only to the agent except its own directories, it has a CPU quota, a lowered priority, and an OOM score that makes the kernel kill the agent before any of your services. Baseline mode limits it to host metrics, discovery and heartbeats and refuses every probe. The source is public if you want to read what runs.

What happens when the agent cannot reach the platform?​

Metrics and heartbeats are spooled to disk and replayed when the connection returns, so the charts fill in. The host is reported down after about a minute of silence, because an unreachable host and a dead one look the same from the outside; the incident resolves on its own when the link is back. A host installed during an outage registers itself when the endpoint answers. See Connectivity.

Why is a rule missing from the rules tab?​

The tab shows only rules whose metric flows in your workspace. A rule about RabbitMQ does not exist for you until a broker appears; then it is there, enabled, with its default threshold. The complete list is the rules catalog.

I resolved an incident and it reopened. Why?​

Because the condition was still true. Resolving is a statement that the problem is gone; when the metric still breaches, the platform reopens the incident within minutes and the timeline shows both your resolution and the reopen. Acknowledge says "someone is on it" and stops the paging without claiming anything. For planned work, open a maintenance window.

The problem is fixed but the incident is still open​

Incidents close when the metric crosses the recovery threshold, which sits below the opening one, and the next sample has to arrive first. Most rules close within a minute or two of the fix. A rule whose source stopped sending entirely (a service that was removed) closes when the metric has been absent for a while.

Why did my host show "down" for a minute during an update?​

It did not, normally: the platform knows which hosts it just asked to update, and the reconnect a few seconds later is not treated as an outage. A host that restarts for other reasons (a reboot, a crash) is reported down after two missed heartbeats.

How do I stop log collection?​

Three levers: switch a service off on the host page (its probes and its logs stop together), toggle Log shipping off in the host's Manage card, or use the workspace-wide switch in Settings. Info and debug lines, including web access logs, are never collected unless you enable full depth for that service. See Logs.

Can I pin the agent version?​

Yes, per host or for the whole workspace, from the host page or the Fleet tab. A pinned host converges to exactly that version, including rolling back, and stays there until you unpin it. See Updates.

How much does the agent cost the host?​

Under one percent of a core and a few megabytes of memory at the production cadence; a busy web server with hundreds of sites uses more during its probe cycles but stays far below what a single PHP worker uses. The numbers and the knobs are in Service management.

Two of my hosts keep swapping names​

They share a machine id, usually because both came from the same image. See Two machines, one identity.

Where is my data stored, and for how long?​

In the EU; the endpoint built into the published agent is European and data is stored and processed there. Retention follows the plan; the windows per plan are on the pricing page, and the mechanics are in Billing and retention. What is transmitted at all is listed in What the agent sends.

Can I remove a host and its data?​

Yes. Remove it on the host page (or run wakora uninstall on the box, which also tells the platform). The host's telemetry is purged and its open incidents are closed. See Uninstall.