sandboxpolicy: size the verdict ring buffers and report drops

Aperta
#1,107 0 commenti 0 reazioni 0 assegnatari Vedi su GitHub

I maintainer di solito rispondono entro 1 giorno

Nessuno ha ancora preso questa issue.

Valutazione

Difficoltà
5/5
Tempo stimato
Più di una settimana
Idoneità per principianti
35/100
Tipo di issue
Funzionalità
Chiarezza
Abbastanza chiara
Stato di attività
Attiva
Stack tecnologico
go, kubernetes

Direzione di ricerca

Start with enforce_file/program.bpf.c and its events ring buffer, then locate the privilege and network gadget buffers and the code that builds policy status and exports metrics. Measure event and drop rates under a file-heavy workload on AL2023 kernels 6.12 and 6.18, and record the buffer budget in the README. Done means drops appear in status and exported metrics, with a test that fills a buffer and verifies the report.

Scritto dal modello di indicizzazione a partire dal testo della issue.

Descrizione

Follow-up from the review of #1075.

Problem

The file gadget sends its verdict events through one ring buffer of 256 KiB (events in enforce_file/program.bpf.c). Under a file-heavy workload the buffer can fill. A full buffer drops the event; the drop is counted (VERDICT_COUNT_DROPPED), but:

  • the size was not derived from a measured event rate;
  • a sustained drop rate is not visible in the policy status;
  • the same question applies to the privilege and network gadgets, which have their own buffers.

Enforcement itself is not affected: the hook decides before it reserves the event. What is lost is the verdict record (audit trail and export).

Proposal

  • Measure the event rate and the drop rate under a file-heavy workload on the fleet kernels (AL2023 6.12 and 6.18).
  • Size each section's buffer from that measurement, with a documented budget per node.
  • Report sustained drops in status (for example a Degraded notice with the drop count), and export the count.
  • Consider aggregating repeated identical allow verdicts in the kernel before they reach the buffer.

Done when

  • Buffer sizes are justified by a measurement recorded in the README.
  • A drop is visible in status and in the exported metrics.
  • A test fills the buffer and checks that the drop is reported.
Lingua principale
Go
Stelle
39
Fork
42
Merge medio
2g 16h
PR unite (30g)
48

Preparare l'ambiente

Come iniziare

  1. Leggi tutta la issue e poi la guida ai contributi del progetto.
  2. Commenta sulla issue per dire che te ne occupi tu — evita che due persone facciano lo stesso lavoro.
  3. Fai un fork del repository e lavora su un branch.
  4. Apri una pull request che faccia riferimento al numero della issue.

Altre issue di kubescape/node-agent

Tutte le issue di kubescape/node-agent

Issue simili

Altre issue su Go

Ricevi le nuove issue nella tua casella

Un breve riepilogo di issue GitHub adatte ai principianti.