trustgraph/trustgraph-base
cybermaggedon b615150624
fix: resolve multiple Kafka backend issues blocking message delivery (#833)
- Producer: add delivery callback to surface send errors instead of
  silently swallowing them, and raise message.max.bytes to 10MB
- Consumer: raise fetch.message.max.bytes to 10MB to match producer,
  tighten session/heartbeat timeouts for fast group joins, and add
  partition assign/revoke logging for diagnostics
- Topic naming: replace colons with dots in topic names since Kafka
  rejects colons (flow:tg:document-load:default was producing invalid
  topic name tg.flow.document-load:default)
- Response consumers: use auto.offset.reset=earliest instead of latest
  so responses published before partition assignment aren't lost
- UNKNOWN_TOPIC_OR_PART: treat as timeout instead of fatal error so
  consumers wait for auto-created topics instead of crashing
- Concurrency: cap consumer workers to 1 for Kafka since topics have
  1 partition — extra consumers trigger rebalance storms that block
  all message delivery
2026-04-18 22:42:24 +01:00
..
trustgraph fix: resolve multiple Kafka backend issues blocking message delivery (#833) 2026-04-18 22:42:24 +01:00
pyproject.toml Add Kafka pub/sub backend (#830) 2026-04-18 11:18:34 +01:00
README.md Maint/fix build env (#84) 2024-09-30 19:47:09 +01:00