jaeger
Catalog date: 2026-06-24.
Distributed tracing backend for the infrastructure trace pipeline —
HTTP, database, queue, and inter-service spans land here. AI / LLM
spans (OpenInference) route to phoenix
instead.
Identity
| Field | Value |
|---|---|
| Service id | jaeger |
| Role | observability |
| Image | jaegertracing/all-in-one:1.76.0 (Kustomize) / jaegertracing/all-in-one:1.55 (compose) |
| Port | 16686 (HTTP — UI/query), 4317 (OTLP gRPC), 4318 (OTLP HTTP), 14268 (HTTP collector), 14250 (gRPC collector), 14269 (admin/metrics) |
| Health | / |
Per the deployment.yaml
and service.yaml
manifests, no 6831 UDP agent port is exposed. There is no split
jaeger-collector / jaeger-query image pair — both are the same
jaegertracing/all-in-one Deployment — but the Kustomize manifest does
front it with two separate Services, jaeger-query (UI + admin-http)
and jaeger-collector (OTLP/collector ports), plus a jaeger-external
LoadBalancer Service for the UI. Correcting the earlier port list and
refining the "split in cloud cells" claim, which is true at the Service
level but not at the image/Deployment level.
Deployment surfaces
| Surface | Where |
|---|---|
| Compose | service jaeger in alphaswarm_platform/compose/docker-compose.yml |
| Kustomize | observability/jaeger/ — single all-in-one Deployment, fronted by jaeger-query / jaeger-collector / jaeger-external Services; no separate cloud-cell manifest was found |
Dependencies
Upstream: otel-collector — fans the infrastructure trace
pipeline here.
Downstream: Grafana (datasource) and the operator's UI for manual span inspection.
Operations
- Storage: the Kustomize manifest sets
SPAN_STORAGE_TYPE=elasticsearchpointed at the in-cluster OpenSearch service (ES_SERVER_URLS=http://opensearch.alphaswarm-observability.svc.cluster.local:9200) — not badger. The compose service does not setSPAN_STORAGE_TYPEexplicitly (Jaeger all-in-one defaults to in-memory storage in that case). No badger configuration was found in either surface — correcting the earlier "in-cell badger / cloud Elasticsearch" claim. - Retention: no retention-policy config was found for Jaeger / OpenSearch in this repo; flagging the earlier "7 days in-cell, 30 days in cloud" figures as unverified rather than restating them.
- Sampling: per the
tail_samplingprocessor incollector-gateway.yaml, theprobabilistic_baselinepolicy issampling_percentage: 1.0(1%, not 5%) plus 100% of error-status spans (and AI/LLM/agent spans matched by their own policies, which route to Phoenix instead — seeotel-collector.md).
See also
otel-collector.md— routing source.observability.md— concept doc.