OVN-Kubernetes ACL drop-log processor — builds flow baselines via Welford's algorithm
2.8K
Part of the NetworkPolicy Intelligence System (NPI).
Polls Loki every 30 seconds for OVN-Kubernetes ACL drop events (collected cluster-wide by a Fluent Bit DaemonSet), parses them into structured flow records, and incrementally builds per-namespace-pair statistical baselines using Welford's online algorithm — no need to store raw history to compute running mean/variance.
Output feeds two downstream consumers:
flow_events — individual flow records (TimescaleDB hypertable)flow_baseline — running statistics per (src_namespace, dst_namespace, dst_port) key, consumed
by the Anomaly Detector for Z-score spike detectionRuns in drops-only mode by default to keep flow volume manageable at scale.
Python 3.11 · Loki HTTP API · psycopg2 · TimescaleDB
docker pull nassimsai/flow-processor:v1.0-pfe2026
Requires: DATABASE_URL env var, network access to the cluster's Loki service.
Content type
Image
Digest
sha256:8b7047d4a…
Size
78 MB
Last updated
4 months ago
docker pull nassimsai/flow-processor