Deploy, Scale and Extend Jaeger
summary
Ticketmaster deploys Jaeger on Kubernetes with collectors and queries in-cluster and Elasticsearch outside it, retaining five days (~2.1B spans, ~1.2 TB) and ingesting about 9,000 spans/sec. Instrumentation covers ~25 services using Java, Node, Go and mobile clients; agents run as sidecars or daemonsets. Multi-region traces use Kafka with a replicator and a separate jaeger-collector consumer. Operational details include index cleanup via cron every five days, scaling collectors to four pods, and storage plugin support in 1.12.
AI-written summary · about
chapters (19)
- 0:00Intro
- 3:27Why tracing
- 4:42Current state
- 6:06How did we start
- 11:28Deploying Jaeger
- 14:32Environments
- 17:05Agent
- 21:28Replication
- 24:31Scaling
- 26:51Logs
- 27:16Current infrastructure
- 27:51Extending Jaeger
- 28:30UI Protection
- 29:52Processor
- 30:11Preprocessor
- 30:54Presafe
- 31:10Dont Fork
- 31:41Analytics
- 32:03Outro
