Skip to content

Clusters

DevEx › Clusters (?tab=namespaces) holds every view of the one cluster as sub-tabs. Since 2026-10-31 it is the merge of the former Cluster, Application consoles and Delivery (GitOps) rows. Every /api/* read here needs a console session.

Above the tabs, one line says how many platform workloads are ready (GET /api/metrics); the list of the ones that are not is Namespaces’. Old links keep working: ?tab=appConsoles, ?tab=metrics, ?tab=instances, ?tab=routing, ?tab=secrets and ?tab=gitops open the sub-tab that now holds their subject, and ?tab=placement opens Nodes & servers.

How to read absence. These endpoints answer 200 even when they could not read a source, and say so in the body (source:"unavailable", complete:false, or an unmeasured[] list with the reason). An empty list means “none”; an unreadable one says it was not read. See Status and provenance.
Sub-tab?tab=ReadsWhat it shows
OverviewappConsolesGET /api/app-consolesthe door grid: one card per tool you may reach, what it is for, whether it is up, how to open it. See Overview
NamespacesnamespacesGET /api/namespacesper namespace: admission posture, workloads with ready/desired, their pods (restarts, node, why the last container died), events, quotas, and one verdict. Container logs are not read.
Nodes & serversmetricsGET /api/cluster-nodes, GET /api/nodes/{name}, GET /api/metrics, GET /api/serverseach node once: conditions, capacity, requests, taints, pods and why each workload landed where it did; then live node usage and the console process; then the servers outside the cluster
InstancesinstancesGET /api/dashboardthe tenant ClientInstances: app, tenant, role, phase, readiness, endpoint, and FinOps attribution; Group by initiative groups them
RoutingroutingGET /api/routeshow traffic reaches each service through the edge, and whether what it reaches is serving
Secrets & PKIsecretsGET /api/securitythe shared mesh CA as distributed to each namespace: fingerprint and expiry
GitOpsgitopsGET /api/gitopsdesired vs running per workload, and drift. See GitOps

Nodes & servers

One tab for every machine; node capacity is drawn once (it used to be on three tabs). Servers are attached dynamically. The off-cluster section lists:

  • the off-cluster servers declared in the tide-servers ConfigMap, such as the inference server that serves the general and coder tiers. Each installed server’s gate /status is read through servefwd; no SSH key is in any pod. What a server serves is the model router’s tiers whose backend is one of its endpoints. A server that does not answer is shown unreachable with the reason, and one declared installed: false is never dialled;
  • any cluster node that serves a model, with what it serves.

Readings are cached for 20 seconds, so several viewers cost one read of the server. An admin can register a server from the last card.

GitOps

Delivery is GitHub Actions: tide-cd.yml builds the images a merge changed, applies the kustomize overlays, rolls the workloads, and records each deploy in the console’s lineage record. There is no Argo CD. Per workload:

  • desired: the newest tide-cd.yml lineage record that planned or rolled the workload: commit, image, digest, run, time and roll verdict;
  • running: what the workload’s pods report (image digests) and the version labels on the live object (app.kubernetes.io/version, runink.org/commit);
  • drift: decided on the digest when the build reported one, else on the commit label. It is null when neither pair could be read, never false.

The newest tide-cd.yml runs are listed beside it. GitOps and Deploy lineage read one record; each links to the other.

A roll pins every container to an image digest. A bare kubectl rollout restart restarts onto the same digest. To move a workload onto new images, run the delivery workflow for that image.

The data-systems namespace

The installation’s data sources (PostgreSQL, MySQL, Kafka and the Kafka Schema Registry) run in one shared namespace, data-systems, not in the namespace of any one application. It is declared by the FACE kit, with default-deny ingress that admits only the data ports from the application and platform namespaces. TIDE registers the three sources from the tide-connection-manifests ConfigMap and reads their credentials from that namespace’s Secrets by reference. It is not a tracked namespace of the Namespaces view, and it has no console page of its own.