Making your Kubernetes-based log collection reliable & durable with
summary
Vector collects Kubernetes pod and node logs as a Rust agent deployed as DaemonSet agents with a three‑container pod: Vector, a reloader, and a metrics proxy, mounting /var/log and a disk buffer. Real incidents: node disks fill because Vector holds rotated files (default in‑memory buffer 1,000 → tuned to 10,000 message buffer or on‑disk buffers), Prometheus metric cardinality explodes from file path labels (expire_metrics_seconds breaks rate() so code was patched), and concurrent API calls spike etcd memory so priority‑and‑fairness + client caching fixes work.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 1:25Why log collection matters
- 5:18What is Vector
- 7:29Vector architecture
- 10:04Log collection topologies
- 12:15Deploying Vector in Kubernetes
- 16:07Outage: Disk space exhaustion
- 19:23Outage: Prometheus explosion
- 22:37Outage: Control plane overload
- 27:10Conclusion
related talks
35mMaking your Kubernetes-based log collection reliable & durable with Vector
25mFoster Secure Innovation with Your Kubernetes Platform
35mDeep Dive: Kubernetes (Instrumentation) SIG
34mShift Left and Security-as-Code: Security on Kubernetes
07mData Gravity and Kubernetes: Managing Large-Scale Data Ingest With Minimal Latency
31mSafeguarding Kubernetes: Pod Security Admission, NeuVector & Kyverno in the Ring
41mBuilding a new Big Data distribution based on Kubernetes with a twist!
17mKeynote: A Cloud Native Swiss Knife