user-authn adds audience:server:client_id:kubernetes to the dex
authenticator's scope only for DexAuthenticators annotated
dexauthenticator.deckhouse.io/allow-access-to-kubernetes, which is what
the console module does. v0.8.0 set sendAuthorizationHeader alone, so the
token in the header was one the API server does not accept.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
After a reboot the registry container exists but is stopped; docker
inspect succeeded, nothing started it, and the readiness loop never ended.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Problems: cards, namespace and node boxes and the tree carry a yellow or
red badge, computed on the server from object state (crash loops, image
pulls, pending and unschedulable pods, readiness older than 2 minutes, a
stuck rollout, a failed CronJob run, Services without endpoints,
Ingresses to missing Services, node conditions), recent Warning events
(notable reasons at once, others after five repeats) and Deckhouse
ClusterAlerts mapped by their labels. A Problems panel lists them all.
Checked against the live cluster: upmeter's synthetic probe pods and
ingress-nginx's fakeservice Ingresses are excluded, they fail on purpose.
Actions: restart, scale, run a CronJob now, suspend/resume, delete a
pod, cordon/uncordon. They run with the viewer's bearer token from the
DexAuthenticator (sendAuthorizationHeader), buttons follow a
SelfSubjectAccessReview, the service account stays read-only. Cross-site
writes are refused by Sec-Fetch-Site/Origin/content type. Scale retries
on conflict: right after a restart a single Get+Update got 409.
Load (metrics.k8s.io against requests/limits/allocatable, per-pod table)
and an Events tab. Cards are locked until the padlock is opened; the map
exports to PNG; the page refreshes badges every minute in place.
Performance, measured on the live cluster:
- layout of big views 3-4x faster: equal-size cards with no edges are
laid out on the grid ELK would produce (same sizes, checked against
ELK on a frozen node view), skipping ~8 ms per ELK call;
- hubble-relay gets ~20% fewer flows: the to-overlay copy of pod-to-pod
traffic is dropped at the agents; 331/331 directions unchanged;
- static files carry ETags, a reload revalidates instead of pulling
~2 MB of vendored scripts.
KUBECONFIG is honoured for running the binary on a workstation.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The flows API sent "edges": null for a window without connections (a nil
Go slice); the frontend iterated it and renderMap threw after the new
breadcrumbs were drawn, so the previous namespace stayed on screen under
the new name. The API now sends empty lists, fetchFlows guards against
null, and a failed draw reports itself in the status bar.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
helm_lib_prometheus_rules only renders when "operator-prometheus-crd" is
an enabled module; on DKP 1.76 that module doesn't exist (the CRDs come
with operator-prometheus), so the two alerts were silently dropped on a
live install. Gate on the PrometheusRule API instead.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
werf keeps every stage in --repo, base images included; pushing
Deckhouse's builder/golang-alpine (762 MB layer) to Gitea over a home
uplink timed out. Stages now live in a local registry:2 and publish
crane-copies the app image, bundle and release, keeping digests.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
.werf/stages/base-images.yaml defines a werf image for every key of
base_images.yml, and Deckhouse's published file has ~1200 keys; a full
build started pulling and storing all of them.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Layout, build and templates follow deckhouse/modules-template; dmt lint
passes with no findings (Go files are excluded from the Flant-only
license-header rule, see .dmtlint.yaml).
- module.yaml, openapi schemas with Russian docs, lib-helm templates:
https modes, ingress class, publicDomainTemplate, node placement,
priority class, PSS-restricted security context, VPA, PDB, Ingress
plus Gateway API ListenerSet/HTTPRoute, DexAuthenticator.
- module-sdk hooks: hubble CA from d8-cni-cilium (public certificate
only), external_auth, copy_custom_certificate.
- The app moves to images/cluster-map/src and is split into packages;
traffic mode Continuous/OnDemand/Disabled and retention are module
settings; graceful shutdown; Prometheus metrics and two alerts;
frontend embedded, English and Russian, strict CSP.
- Built with werf on Deckhouse's published base images; Makefile
publishes the registry layout modules-actions uses.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Host-originated flows, remote-node, cilium health and DNS :53 were ~74%
of the non-reply events on a 12k-event sample and none of them map to a
workload card; streaming them all kept hubble-relay at ~400m CPU. They
are now blacklisted in the GetFlows request, so agents drop them before
the relay. The UI's "Служебный" toggle had nothing left to show and is
gone; the legend says what isn't collected.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The ring buffer on each agent holds 4095 flows, ~10 s at the observed
~380 flows/s, so the 5/15/60 min windows were really showing the last
few seconds. A background follow stream now fills per-minute buckets
for the last hour, and counts are distinct connections (5-tuples)
instead of Hubble events, with opened (SYN) and dropped broken out.
/api/flows reports coveredSec so the UI can say how much of the window
the collector has seen since start.
/api/graph reads informer caches instead of listing every kind per
request (~0.9 s before). Pod-to-workload resolution for flows uses the
same cache. YAML view drops managedFields.
Frontend: traffic refreshes in place every 30 s without relayout; dash
animation rationed (12 fps, only highlighted edges in dense views,
paused after a minute idle) since each step redraws the whole element
layer, ~15 ms on a 200-element view.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Frontend rewritten around three panes. Inventory tree on the left
(namespaces by kind, or nodes by namespace) with search and keyboard
navigation replaces the namespace checkbox list. The map shows only the
selected namespace(s) or node: fixed-size cards inside namespace boxes,
laid out in two elk passes so a big namespace packs into rows instead of
one tall column; zoom is capped so the view no longer jumps. Details
panel has Overview / Links / YAML / Logs tabs. Light and dark themes.
Libraries are vendored, no CDN.
Backend: /api/flows reads recent flows from hubble-relay (TLS verified
with the hubble CA, no client cert needed) and aggregates them per
source/destination/port. Pods are resolved to workloads with the same
rules as the graph, since Hubble leaves Workloads empty on some flows.
Pods under non-native controllers (kruise DaemonSet, CNPG clusters,
bare Jobs, static pods) are now graph nodes too. Probes move to
/healthz: /api/graph was timing out the 1s probe.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Deployment/StatefulSet/DaemonSet/CronJob details now show a pod +
container picker with the pod's recent log tail (500 lines, no follow),
sitting alongside the existing raw-manifest view. Pod resolution reuses
buildGraph's ownerReferences-walking logic (ownerWorkloadID) so it stays
consistent with the schedules-on edges already shown on the graph.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>