phoenixai-kubernetes-operator

PhoenixAI Anywhere Console — UI guide

A high-level tour of the Anywhere web console: what it does, who signs in, where each user goes, and what every page shows.

1. What the console does

PhoenixAI Anywhere is the self-hosted console for PhoenixAI (StarRocks enterprise) clusters managed by the PhoenixAI operator on Kubernetes. It is a read-only console: it observes and reports, and never modifies cluster data.

Its capabilities fall into four areas:

Area	What you get
Insight	Cluster inventory and health, node/pod topology, CPU / memory / disk monitoring, query insights, audit-log search
License	Licenses registered on each cluster, effective-license rules, expiry and capacity tracking, license registration
Metering	CCU usage per cluster and per day, metering periods, and a full auditable record set with downloadable raw records
Support	Support bundles — a scenario-driven collector that packages manifests, configs, logs, SQL output and diagnostics for support

Two supporting capabilities appear inside those pages rather than as destinations of their own: saved SQL commands (read-only SELECT/SHOW commands per cluster) and shell commands (saved commands run inside a pod, for example jstack or disk usage).

Only shared-data (“elastic”) clusters are supported. A shared-nothing (“classic”) cluster may still be listed with its type reported, but feature pages do not operate on it.

2. The two kinds of users

The console is two consoles side by side, each with its own sign-in and its own account model.

Admin Console

The platform operator’s view. Covers every cluster the installation can see, plus the install-level pages (usage, license, health checks, support).

• Admin accounts live in a Kubernetes Secret, managed with kubectl or Helm at install time. Several admin accounts may exist; they are added, removed and rotated by editing that Secret.
• Admins are not database users and need no database credential. For cluster SQL, Anywhere resolves the cluster’s own root identity from the cluster CR’s Secret automatically — no FE address or password is ever configurable in the console.
• In this release all admins are equivalent: any successful admin sign-in has full access to every admin page.

Cluster Console

A single cluster’s own users. Covers data catalog browsing, query insights and system monitoring for the one cluster the session is bound to.

• Accounts are PhoenixAI database users, defined inside each individual cluster. Anywhere holds no user store for them — no create, delete or password reset; they are managed in the cluster with SQL (CREATE USER, GRANT, …).
• Sign-in selects a cluster, then takes that cluster’s database username and password. Anywhere validates them by opening a real connection to the cluster’s FE, so a rejected credential is a failed sign-in, and an unreachable FE is reported as a distinct error.
• The session is bound to that one cluster and passes the credentials through on every query, so what a user can see is decided entirely by the cluster’s own SQL privileges, not by Anywhere.
• Accounts are per-cluster: the same username in two clusters is two unrelated accounts with independent passwords. That is why the sign-in form asks for the cluster first.
• root is PhoenixAI’s built-in superuser. Any user that can connect to the FE can sign in — but root is not merely a convenient example: the Audit logs tab is visible only to root, as §5 describes under Query insights.

:::caution Credentials are passed through, so changing a password ends the session Because every query carries the credentials the session was opened with, changing that user’s password in the cluster invalidates the running session immediately. The next page load reports cluster credentials were rejected; please log in again.

Two things are worth knowing about that state. First, the Session expired dialog it raises says “You will be redirected to sign in again” but does not redirect, and it blocks the page underneath — including the user menu, so there is no way to sign out from it. Navigating to /cluster-console/login returns you to the same dialog; clearing the session cookie is the way out. Second, if you change root’s password, you also break Anywhere’s own cluster-usage path, because Anywhere authenticates as root from the cluster CR’s Secret — query-history collection starts logging access denied until the two agree again.

:::

How the two interact

The two sessions are independent and coexist in one browser: signing into one console never signs you out of the other, and each keeps its own cookie. They do not share access either — admin pages are not reachable with a cluster session, and a signed-in cluster user who opens an admin URL is redirected to the admin sign-in page.

The Cluster Console sign-in shows a hint when a cluster session is already active, naming the identity being replaced. An active admin session does not raise that hint — the two are separate cookies, and this form only probes the cluster one.

3. Where each user goes

Both consoles are served by the same installation, on the same host and port, distinguished by path.

Admin Console paths

Page	Path
Sign in	/login
Clusters (landing page; / redirects here)	/clusters — table view at ?layout=table
Cluster detail	/clusters/{namespace}/{name} — tabs at ?tab=
Warehouse detail	/clusters/{namespace}/{name}/warehouses/{warehouse}
Usage & Metering	/usage — tabs at ?tab=
License	/license
Health checks	/health-checks
Support	/support
Create support bundle	/support/create

The sidebar carries four destinations — Clusters, Usage & Metering, Health checks, Support — and can be collapsed to icons. License has no sidebar entry: reach it from a cluster’s License tab, or by URL. Below the destinations sit a Help document link and the signed-in user’s menu, which holds a Theme selector (Light / Dark / System), Sign out, and the running version.

Cluster Console paths

Page	Path
Sign in	/cluster-console/login
Data catalog	/cluster-console/catalog
Databases in a catalog	/cluster-console/catalog/{catalog}
Database detail	/cluster-console/catalog/{catalog}/{database} — tabs at ?view=, job kind at ?task=
Table detail	/cluster-console/catalog/{catalog}/{database}/tables/{table} — tabs at ?tab=
Query insights	/cluster-console/query-insights — audit sub-tab at ?view=audit
Query detail	/cluster-console/query-insights/{queryId}
System monitoring	/cluster-console/monitoring

Cluster Console URLs never name the cluster — the session supplies it. That is why the paths are the same for every cluster, and why switching clusters means signing in again. The signed-in cluster is shown in the header (phoenixai/kube-anywhere in the screenshots below), beside the three destinations and the user menu.

Pages that render times take a display timezone from the UI and put it in the URL (?tz=UTC), so a link reproduces what the sender was looking at.

Finding the other console. Both paths render the same sign-in page, which carries two tabs — Platform administrator and Database user — so neither audience has to be told a URL: the path you arrive on simply preselects its tab. There is also a hand-off from a cluster’s detail page in the Admin Console (Open cluster), which opens that page on the Database user tab with the cluster preselected; that is context only, credentials are still required.

The cluster picker is preselected from ?cluster=ns/name (the admin hand-off, or a bounce carrying an expired session’s cluster) or, failing that, from the last cluster signed into from that browser. It is not preselected just because the installation has only one cluster.

4. Admin Console pages

What a brand-new environment looks like. Several pages are empty until time passes or a setting is turned on, and that is normal rather than broken: Usage & Metering reads all zeros (Clusters tracked 0, Last metered at —) because usage is sampled on an interval; the cluster Overview shows — for month-to-date CCU and “No data for this metric” for data volume, because that comparison needs a 24-hour baseline the cluster is too young to have; Support has no bundles until someone creates one; and Query insights is switched off by default — see that section for the two settings it needs.

Sign in — /login

Username and password against the admin Secret, beside a product hero panel. A session-expiry notice appears here when an earlier session timed out, and sign-in returns you to the page you were trying to reach (carried as ?redirect=). A theme toggle sits in the top-right corner.

Clusters — /clusters

The landing page: every cluster the installation can see, in card view or table view (?layout=table). Both show name, status (Running / Creating / Stopped / Failed / Unknown), type, version, coordinator address, warehouse count, namespace, creation time and license expiry — with a near-expiry license highlighted. Selecting a cluster opens its detail page.

Cluster detail — /clusters/{ns}/{name}

One cluster, as seven tabs addressed by ?tab= so links stay shallow (overview is the default and carries no parameter). Open cluster in the top-right hands off to the Cluster Console for this cluster.

Overview — cluster properties (name, namespace, type, version, status, warehouses, created, license expiry, time zone), month-to-date CCU, node and warehouse activity, data-volume change over 24 hours, and a 7-day cluster usage trend, annotated All times in UTC.

Warehouses — one card per warehouse including the built-in one, badged Builtin warehouse or Normal warehouse, with requests and limits, replicas, autoscaling range, creation time, pods running/total and utilization.

Resources (breadcrumb: Resources & Topology) — a health strip (nodes ready, warehouse count, “All components healthy”), then the cluster object itself and expandable groups: Coordinator (shared across all warehouses) and Warehouses, each warehouse running its own compute nodes, with each group naming its image and its requests and limits. Each row’s View shows its manifest and recent Kubernetes events; pods carry live CPU and memory against requests.

Monitoring — charts over a selectable time range and display timezone, split between Cluster state and Instance state, with a one- or two-charts-per-row layout that can be reordered, and a per-chart aggregation selector. The display timezone defaults to UTC+00:00. Which charts appear, and which aggregation each defaults to, depends on what the cluster’s PhoenixAI version exposes — treat any specific set as an example rather than a fixed list; here it is Cluster Data Size, Query QPS, Ingested Times and Ingested Rows. Each chart title carries a tooltip describing what that chart measures. An empty chart means no data in the selected window; if one is empty when you expect data, run the Prometheus dependency check described in Point the Anywhere Console at Prometheus.

Health checks — this cluster’s checks by type (info / warning / critical), status, component and message, each naming the check that produced it, with sortable columns, filters and pagination. A real cluster produces a substantial list — a couple of dozen findings on a newly built cluster — and they are specific and actionable rather than generic: an expiring license, CPU limits set on latency-sensitive components, multi-replica components not spread across topology domains.

Diagnostics — two collapsed libraries, each reporting how many entries it holds (16 SQL commands and 1 shell command here). SQL commands runs saved read-only commands for this cluster; Shell commands runs a saved shell command inside one pod. Expanding a library reveals its picker, Run, and New command; results report exit code, stdout and stderr. Creating custom shell commands is gated by configuration and can be disabled entirely.

License — the effective license for this cluster: what it is bound to, expiry with days remaining, capacity used against licensed cores, Register license, and the registered-licenses table marking which one is effective. Times are UTC.

A newly created cluster already carries a license — nobody registers one by hand — but it is short-lived, so a fresh environment shows Expiring soon and Health checks raises license-invalid as critical (until 2026-08-24 this rule was license-expiring-soon, before it merged with license-missing). That is expected on a new cluster, not a misconfiguration.

Warehouse detail — /clusters/{ns}/{name}/warehouses/{warehouse}

A warehouse as its own page rather than a dialog, so it is deep-linkable, with its own Overview, Resources and Health checks tabs. Overview carries requests and limits, replicas, autoscaling, created, cluster, version, pods, utilization and image.

Usage & Metering — /usage

Install-level metering across all clusters, in three tabs. All times are UTC. Clusters running a PhoenixAI version that does not expose the usage metric are called out as not metered, with the required version stated, rather than shown as zero.

Overview — this month’s CCU, clusters tracked, collector status, last metered at, and the metering periods table with the CCU formula in force (CCU = 0.9 x coreSeconds/3600/4 (v1)).

Usage — the cluster usage graph per day, filterable to one cluster or all, over 1M / 3M / 6M / YTD / 1Y / MAX, switchable between Graph and Details.

Audit — the complete record set behind those numbers: record count, sampling interval, how CCU is derived, per-cluster first/last record and sequence range, counter resets explained, and Download JSON of every raw record with sequence numbers and hashes for verification.

License — /license

The install-level license page, with a cluster selector in the header (licenses are cluster-scoped) and Register license. Shows what the license is bound to, expiry with days remaining, and capacity used. Times are UTC.

:::caution License history does not load in this release The License history panel — activation, renewal and alert events — is served by an endpoint the backend has not built yet (/api/v1/license/events, still marked a proposal in the frontend’s endpoint map). Against a real installation the request falls through to the SPA and the panel reports Failed to load … is not valid JSON. The three summary cards above it are real.

Per-cluster license facts, including the registered-licenses table, are on the cluster’s own License tab and do work. :::

Health checks — /health-checks

Every check across all clusters in one table — cluster, type, status, component, message and check id — filterable by cluster, type, status and component, with sortable columns and pagination. Findings are ordered by severity, then component, then rule id.

Support — /support

Support bundles for the selected cluster: bundle name, cluster, created, size, status and the categories collected. Finished bundles download from the name; the row actions view details, copy an export link, and delete. An Interrupted bundle (Anywhere restarted mid-run) offers resume instead. Sensitive values are excluded from bundles.

Create support bundle — /support/create

A full-page composer, which collapses the sidebar to icons. The Scenarios rail on the left (“pick what you’re seeing and we’ll tick the artifacts support usually needs”) preselects a set — compute node crash, query trace, heavy or failing SQL, coordinator stall — or Something else to choose freely; each names how many artifacts it selects. On the right, What to collect lists every artifact with what it is for, its collector id (cluster-info, config, logs, commands, jstack, pstack, jvm-profile, audit-analysis, cn-crash-context, metrics-snapshot, cn-memory), a shared time window, and per-artifact parameters. Log collection can take whole files or only lines matching a search — the choice that decides whether hundreds of megabytes or tens are transferred. The page warns when a selection is incomplete and an artifact would be silently skipped.

The footer carries an estimate — size and time to collect and download, with Recalculate — and one primary action, Create bundle. A large bundle asks for confirmation first.

Not found — /404

A full-bleed error page with a link home. Unknown URLs resolve here without a session, so a typo or stale link never shows a sign-in page.

5. Cluster Console pages

Sign in — /cluster-console/login

The Database user tab of the shared sign-in page. Select the cluster from the list, then enter that cluster’s database username and password. The two credential fields are labelled simply Username and Password — the same as the administrator tab, because both tabs ask for the same kinds of thing. The form notes that cluster credentials come from your platform admin. Reaching the administrator sign-in is a tab away, and the brand hero panel is shared by both.

Data catalog — /cluster-console/catalog

Read-only browsing of the signed-in cluster, drilling down through searchable levels.

Catalogs — every catalog with its type and comment. A cluster with no external catalogs configured shows only default_catalog, typed Internal; external catalogs such as Hive appear here as they are added.

Databases — the databases in a catalog, with table count and size.

Database detail — tabs for Tables, Materialized views, Views and Tasks (?view=). The Tables tab lists rows, size, creation time, and the table model under a column headed Remark (OLAP for an ordinary table) rather than a free-text description.

Tasks groups jobs by kind in a left-hand rail — Kafka Import, Other Import, Export, Schema Change, Creating View (?task=) — and the columns differ by kind: an import job carries its identifying Tag, state, table, created, paused, task count, progress and the reason a stage changed, and is searchable by job name; a schema-change job carries table, status, progress, start and end time, timeout and details, and is searchable by table. A kind with nothing to show reports “No tasks reported.”

Table detail — tabs for Columns (type, nullability, key, default, extra, comment), DDL and Partitions (?tab=).

Partitions lists state, visible version, partition key, range, bucket column, buckets, duplications and size, one row per partition. In shared-data mode the Duplications column reads 0 — replication is the object store’s job, not the cluster’s — and an unpartitioned table shows a single row named after the table.

Query insights — /cluster-console/query-insights

Query records with filters: status, time range (Last 3 hours by default) and display timezone, user, duration bounds, and free-text search over SQL or query ID, plus a Columns control. Columns follow the familiar query-history vocabulary — start time, query ID, status, SQL, warehouse, duration, memory, client IP, query user, source, queue time, CPU cost, scanned bytes and rows. The list pages by scrolling and reports when it reaches the end.

Records are scoped by the signed-in user’s own SQL privileges, so root sees the cluster’s internal statistics queries and other users’ work, while an ordinary user sees only what it is entitled to.

:::note Audit logs is visible only to root Audit logs is a sibling sub-tab of Query insights (?view=audit), but it is rendered only when the signed-in cluster user is root. Any other user — however privileged in SQL — sees Query insights with no sub-tabs at all and no indication that the tab exists.

That matters on a default install: nothing sets a root password, so Anywhere authenticates as root with an empty one. Reaching Audit logs means being able to sign in as root at this form, and giving root a password to do so is what breaks Anywhere’s own root access (see the caution in §2).

:::

The page has two distinct empty states, and each names its own remedy:

• Query collection is off on this cluster. This is the default, so it is what a new installation sees. Two settings in two different places must be enabled — queryHistory.enabled in the Helm values, and the FE configuration enable_collect_query_detail_info on every FE. The banner names both.
• Collection is on, but not every query is kept. The page then states the policy in force — for example “showing slow queries ≥ 5000 ms and failed queries only” — instead of appearing empty for no reason. A separate banner flags when query profiles are switched off on the cluster, and what to set to enable them; that banner can be dismissed for the visit.

The time-range control offers quick ranges from the last 5 minutes to the last 14 days, plus a custom range with a calendar and from/to times. Changes take effect when Search is pressed, and an absolute range keeps the instants it was given when the display timezone changes.

Audit logs — pick an FE node and a time window, optionally a keyword or query ID, then search. Results show start time, query ID, state, cost, user, client address and statement. Statements are recorded as the FE saw them, with credentials masked.

Query detail — /cluster-console/query-insights/{queryId}

One query record, as Query overview and Query profile. Overview carries the summary (user, database, catalog, warehouse, FE, client IP, timings, queue time, CPU cost, memory, scanned rows and bytes, digest, source), the full SQL with a Format action, and an Explain section.

Query profile is the cluster’s own profile text — summary, planner timings and the operator tree — in a scrollable reader with a copy action. It is present only for queries the cluster profiled, and where it is absent the tab says so on the record itself, naming the query’s latency and the big_query_profile_threshold that excluded it. A tutorial-sized query finishing in milliseconds is well under the default 30s threshold, so “not profiled” is the ordinary answer there rather than a sign that anything is wrong.

System monitoring — /cluster-console/monitoring

Readings for the cluster, over a selectable range and display timezone (UTC+00:00 by default), with a note stating the window the “current” figures cover.

Cluster state — active and queued query counts, plus cluster-wide charts (Cluster Data Size, Query QPS, Ingested Times, Ingested Rows). A metric the cluster does not expose says so in place of the number — query queue metrics, for instance, report as unavailable rather than as zero.

Instance state — a node tree (coordinator nodes, and compute nodes grouped by warehouse) that scopes the charts to the selected instances, each node listed with its address and pod name. Charts here cover FE JVM heap used and utilization and, when the Prometheus behind Anywhere also scrapes kubelet/cAdvisor, pod CPU and memory utilization. Charts and Latest values are switchable views of the same series.

6. Conventions worth knowing

• Time zones. Usage & Metering and License render in UTC and say so. Every other page renders in browser-local time with no zone label. Values the cluster itself formatted are shown verbatim and annotated with the cluster’s own time zone (for example “times in cluster time_zone: UTC” above the catalog tables), because they carry no zone of their own. Monitoring and Query insights offer a display-timezone selector, which affects display only and defaults to UTC; charts state the zone they are drawn in.
• Themes. Both consoles carry a Light / Dark / System selector — in the user menu once signed in, and as an icon on each sign-in page.
• Read-only. Nothing in the console modifies cluster data. Only catalog-listed SQL is ever executed, and execution requests name a saved command rather than carrying SQL.
• Custom shell commands are gated by configuration. When disabled, built-in read-only commands remain available and the console says custom presets are off.
• Anywhere’s own queries are visible. Anywhere tags every statement it runs with /* APP=PhoenixAI Anywhere */. With queryHistory.slowQueryMs at its default those short queries are never persisted, but set it to 0 to keep everything and the console’s own polling appears in Query insights alongside user queries.

7. Known gaps in this release

What	Where	Detail
License history never loads	/license	Backed by an unbuilt endpoint; the panel shows a JSON parse error
Audit logs hidden from non-root users	/cluster-console/query-insights	Gated on the username being exactly root, with no hint for anyone else
Session expired dialog is a dead end after a credential rejection	Cluster Console	When the signed-in user’s password changed in the cluster, the dialog says it will redirect, does not, and blocks the page under it; only clearing the session cookie escapes. A plain session expiry redirects normally