Connected and Disconnected
Status is not a stored flag — it’s derived live from whether the plane is actively connected right now. A plane that has stopped, crashed, or was never deployed reads Disconnected honestly; nothing here is cached in a way that could go stale in the other direction. A freshly registered plane starts Disconnected and stays that way until you deploy it with the credentials you were shown at registration and it connects. That’s expected — see Register a data plane.Why does a plane read Disconnected when it shouldn’t?
Check, in order:- Is the deployment actually running? If it crashed or was never deployed, that’s the fix. A
worker that cannot read or update the
altimate-apps-namespacesConfigMap in the plane’s namespace retries for about 30 seconds and then stops, and its log says “refusing to start” and namesaltimate-apps-namespaces. The ConfigMap records every namespace data apps ran in, so that deleting a tenant also removes its apps. The worker needs permission to get, create and update ConfigMaps in its own namespace. The plane’s install grants that, so this usually means the cluster’s API was unreachable or someone edited the worker’s Role. Once the cause is fixed, the worker starts on its next restart. - Is it configured for the right plane? Each plane’s registration secrets are unique to that plane and are shown once, at registration. A deployment configured with another plane’s secrets never produces an error — it never receives work meant for this plane, so the plane sits Disconnected even though the deployment looks healthy on your side. This is the single most common cause of a plane that looks fine everywhere except the console.
- Does it hold the current plane credential? If you rotated it recently, redeploy with the new value — see Register or join a plane.
How do I send support the details of a plane in my cluster?
If the plane runs in your own cluster and still reads Disconnected, or a pod will not start, collect a support bundle and attach it to your support request:Collect a support bundle
.tar.gz file with your plane’s pod states, recent logs, install values,
events and network policies, with secrets redacted. Read it before you send it. See
dip plane support-bundle for what it collects.
Recent errors
Recent errors are failures a data plane reports about itself: things that went wrong on the plane while it stayed connected, so that Connected alone would hide them. The plane sends them each time it connects, about once a minute, and the console shows the ones from the last 24 hours. Click the recent errors badge in the Status column to see each one: what kind of failure it is, which part of the plane reported it (and which copy of it, when the plane runs several), how many times it happened, when it last happened, and the error message.
An error stays in the list until it has not happened for 24 hours, even if the part of the plane
that reported it restarts in the meantime, and then clears on its own. The count is at least the
number of times it happened; a restart starts counting again from the restart.
Messages never contain secrets: credentials, keys, tokens, passwords, signed links and email
addresses are replaced with
[redacted] before the plane sends them, and again when your tenant
receives them. A message is at most 240 characters, each copy of each part of the plane reports at most eight
errors, and the console shows at most 64 errors per plane.
Who sees recent errors?
Only the tenant that owns the plane. On a plane that several tenants share, the errors can be about any tenant on it, so a tenant that joined someone else’s plane sees “This plane reported errors that only its owner can see.” instead of the list. Failures the plane cannot observe itself, such as a pod that cannot pull its image, do not appear here. Check the plane’s own cluster for those.Plane versions
A plane version is the release a data plane runs, which the plane reports each time it connects. Your tenant’s control plane supports planes a few releases behind it, so you can upgrade a plane in your own maintenance window instead of on every release. A plane stays supported for 90 days after a newer plane release that changes how planes and the control plane talk to each other, or while it is at most two such changes behind, whichever lasts longer. Most releases change nothing a plane depends on, so most releases do not start that clock. The Version column shows:
A plane installed before version reporting existed reports no version. It is treated as one change
behind from the day reporting was added, so it keeps working and is warned before its window closes,
like any other plane that is behind.
From the CLI
List your data planes
errors line, for example:
errors line says they were withheld.
With --json, each plane also carries recent_errors (each with component, instance, kind,
message, count, first_at and last_at) and recent_errors_withheld.
With --json, each plane carries chart_version, contract_version, version_status
(current, supported, warn, refused, ahead or unreported), version_message and
upgrade_by.