Collect Hubble traffic continuously; graph from informer caches

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>
This commit is contained in:
Vitaliy Chefanov
2026-09-23 20:23:20 +03:00
co-authored by Claude Opus 5.5
parent cf84e9976f
commit 6ead0a0d6d
5 changed files with 547 additions and 237 deletions
+25 -19
View File
@@ -10,7 +10,8 @@ Topology map of the Project Chimera Deckhouse cluster.
→ workload by label selector); dashed animated arrows are real traffic
from Hubble Relay over the chosen window (5 min / 15 min / 1 h), red when
Cilium dropped some of it. Peers from other namespaces are drawn as
dashed "ghost" cards.
dashed "ghost" cards. Traffic refreshes in place every 30 s without
relayout; peers that would need new cards wait for "Обновить".
- **Right**: details for the selected object — overview, links (config and
traffic, clickable), live YAML, pod logs.
@@ -20,18 +21,27 @@ way Hubble names them, so traffic attaches to the same cards.
## Backend
Go, `client-go` in-cluster config. Every `/api/graph` call lists the API
fresh — no cache, no watch. `/api/flows` reads recent flows from
`hubble-relay.d8-cni-cilium.svc:443` (TLS, verified with the public part of
Deckhouse's hubble CA mounted at `/etc/hubble-ca/ca.crt`; relay does not
ask for a client certificate) and aggregates them per source → destination
→ port. Replies are excluded, so each connection counts once.
Go, `client-go` in-cluster config. The graph is built from informer caches
(watch-backed, managedFields stripped), so `/api/graph` costs milliseconds
instead of ~0.9 s of List calls.
Traffic: each Cilium agent keeps only 4095 flows in its ring buffer — at a
few hundred flows/s that is ~10 s of history, so reading the buffer on
request can't answer "last 15 minutes". A background follow stream from
`hubble-relay.d8-cni-cilium.svc:443` (TLS verified with the public part of
Deckhouse's hubble CA at `/etc/hubble-ca/ca.crt`; relay does not ask for a
client certificate) feeds per-minute buckets for the last hour. Counts are
**distinct connections** (5-tuples), not Hubble events — one connection
produces several events. `new` = opened by TCP SYN inside the window,
`dropped` = tuples with a DROPPED verdict. After a restart the collector
only knows what it has seen since; `coveredSec` says how much of the
window that is.
| Endpoint | What |
|---|---|
| `/api/graph` | nodes + config edges + pod placement |
| `/api/flows?namespaces=a,b&window=300` | aggregated Hubble traffic |
| `/api/resource?kind=&namespace=&name=` | live YAML |
| `/api/flows?namespaces=a,b&window=300` | aggregated traffic touching those namespaces |
| `/api/resource?kind=&namespace=&name=` | live YAML (managedFields stripped) |
| `/api/pods?id=wl/<ns>/<kind>/<name>&namespace=` | pods of a workload |
| `/api/logs?namespace=&pod=&container=` | last 500 log lines |
| `/healthz` | liveness/readiness |
@@ -46,16 +56,12 @@ CDN. Theme follows `prefers-color-scheme`.
## Build and deploy
The in-cluster registry can't be fed from inside the cluster yet (Kaniko's
image lives on gcr.io, blocked here), so for now:
```
docker build -t cluster-map:vX .
docker save cluster-map:vX -o cluster-map.tar
scp cluster-map.tar xeon:/tmp/
ssh xeon 'sudo /opt/deckhouse/bin/ctr -n k8s.io images import /tmp/cluster-map.tar'
R=git.project-chimera.ru/fanchji/cluster-map/cluster-map-app
docker build -t $R:vX . && docker push $R:vX
docker inspect --format '{{index .RepoDigests 0}}' $R:vX
```
The Deployment pins `nodeSelector: xeon` + `imagePullPolicy: Never`.
Manifests: `03-kubernetes/apps/cluster-map/` in `project-chimera`, mirrored
to `k8s-manifests` for Argo.
The Deployment pins the image by digest. Anonymous pull works (public Gitea
user), no imagePullSecret. Manifests: `03-kubernetes/apps/cluster-map/` in
`project-chimera`, mirrored to `k8s-manifests` for Argo.