Platform services

Use Admin > Platform > Services on a self-managed installation to see the state of every service the installation runs, restart or stop one service, and follow the maintenance job the action starts. Use it when a service looks unhealthy, when something about a service changed on the host, or when you want to know what a restart is doing.

Read the service state

1. Open Admin > Platform, and click the Services tab.
2. Read the strip above the groups. It counts the services that are up, for example 8 / 9 services up, counts how many one-shot setup services completed when the installation has any, for example 2 / 2 init completed, and flags each condition that needs attention with its count, such as 1 unhealthy. Click Refresh to reload the state at once; the panel also keeps itself current.
3. Open a group to see its services. Services are grouped by tier: Edge, Core, AI services, Data tier, Reporting and Observability. A group that holds a service needing attention opens on its own and cannot be collapsed until nothing in it needs attention.
4. Read each row:
   • The service name and, on hover, its description. The critical badge marks a service the platform depends on; the one-shot badge marks a setup service that runs once and exits.
   • The condition badge: Up, Completed, Starting, In progress, Unhealthy, Flapping, Failed, Down or Unknown.
   • The container's name, or no container when the service has none here.
   • A restart count beside the badge, a ↻ and a number, shown only when it is above zero; hovering it reads Docker restart count.
   • The container's own status line, for example Up 3 hours (healthy), with an error after it when the container has one, or Not deployed when the service has no container here.

Finished one-shot setup services are hidden by default. Turn on Show completed init in the strip to bring them back; while they are hidden, the group header counts how many.

Restart or stop a service

1. Open Admin > Platform > Services.
2. On the service's row, click the restart icon or the stop icon. Each carries the service name, for example Restart <name>.
3. Read the confirmation before you confirm:
   • The title names the action and the service, for example Restart <name>?.
   • A restart brings the container back in place. Services that depend on it keep running and reconnect on their own, and a restart does not apply configuration changes.
   • A stop takes the container down, and it stays down until an update or someone on the host starts it again. The Services tab has no start button.
   • For a critical service, the confirm button reads Restart anyway or Stop anyway; otherwise it reads Restart or Stop.
4. Confirm. A toast reads Restart started or Stop started with the service name, the job appears in the Jobs list, and the row usually moves to Starting straight away.

If the action cannot start, the toast reads Could not start the restart or Could not start the stop, with the reason.

Follow the jobs

The Jobs list sits beside the service groups (below them on a narrow window) and shows the installation's maintenance jobs, newest first. Each row carries a status badge, running, completed, failed or cancelled, and a label such as Restart, Stop, Update, Pull images or Apply configuration, with its message.

A job stays in the list for an hour after it ends; when nothing has run yet, the list reads No jobs yet. While a job runs, its live output appears above the list. The output follows the end as new lines arrive; scroll up and it stops following until you return to the bottom. An empty output reads Waiting for output…, and if the connection drops, the corner reads Reconnecting… while the page reconnects on its own.

Permissions

The Platform entry appears when at least one of your roles holds a platform view permission: View Platform Services, View Container Logs, View Sensitive Container Logs or Check For Platform Updates.

The list shows only the services your role may see; services you may not see are left out, and the counts describe what you can see. An action on a service your role cannot touch is refused as a whole: a restart is refused with Access denied. You do not have permission to execute the service and the service name, and a stop with Access denied. You do not have permission to delete the service and the name.

Platform Administrator holds the platform permissions; Platform Viewer is read-only. See Roles and permissions and the Permission reference.

On a managed installation

If DataMind manages your installation, Services and Updates are hidden and only Logs remains; the page shows the version this installation runs and the line Services and updates are managed by DataMind. There is nothing here to restart or stop. See Platform for the area and who can open it.

Limits and edge cases

• The installation handles one maintenance job at a time. While a job runs, restart and stop wait for it to finish, and the row buttons stay disabled.
• A restart or stop request carries between 1 and 50 service names, each at most 200 characters; from this panel you act on one service at a time.
• While the Installer is replaced, service controls are unavailable for a few minutes, usually under five; a notice on the panel explains this. See Platform updates and the Installer card.
• The panel keeps itself current and refreshes faster while a service is still settling or a job runs. Click Refresh to reload the state now.
• A row with no container, or a one-shot service, has no restart or stop button. Stop is greyed out while the service is not running.
• While the panel and its jobs load, they read Loading platform services and Loading jobs.
• If the panel cannot load, it keeps its outline and shows No control plane behind the gateway, Control plane can't reach its backend or Control plane not reachable, with Retry. If only the groups fail, a strip reads Service groups unavailable: with the reason and Retry. If the platform cannot read the group definition, the strip reads Service schema unavailable - groups are approximate and every service is shown in one group. When the deployment files on the host are incomplete, the strip reads Deployment files incomplete on the host. See Observability troubleshooting for the messages in this section.
• Your collapsed groups and the Show completed init switch are remembered in this browser.

Questions, answered

Why do I see only some of the installation's services?

The list shows the services your role may see. Services your role cannot see are left out, and the counts describe the visible set. Ask an administrator if you need to see more.

What makes a service critical?

The platform depends on it: restarting or stopping one takes the platform down for everyone. Its confirmation uses stronger wording, Restart anyway or Stop anyway.

What is a one-shot service?

A setup service that runs once and exits, for example one that prepares the database. A finished one is hidden by default; turn on Show completed init to see them, and the group header counts the hidden ones while they are hidden.

Why does a row have no restart or stop button?

Either your role lacks the matching permission, Restart Platform Services or Stop Platform Services, or the service has no container here or is a one-shot. Those rows cannot be restarted or stopped.

Why is there no way to start a service on this page?

The Services tab has no start button. A stopped service stays down until an update or someone on the host starts it again.

Why are the buttons disabled while a job runs?

The installation handles one maintenance job at a time, so restart, stop and updates wait for the running job to finish. The buttons return when it ends, and Stop is greyed out while the service is not running.

What does the live output above the Jobs list show?

It shows the running job's output as it happens, so you can watch a restart, a stop or an update without leaving the panel. The view follows the end while you read; scroll up and it stops following until you return to the bottom.

Why is every service shown in one group?

The platform could not read the definition that groups the services, so it falls back to one group, and the strip reads Service schema unavailable - groups are approximate. The states still come from the running installation; click Refresh to try again.

Why are the Services and Updates tabs missing?

DataMind manages this installation, so the local services and updates are handled for you; the page reads Services and updates are managed by DataMind and only Logs remains. There is nothing for you to do here.