Open durandom opened 3 years ago
Currently Istio reports it's tracing data directly to a backend system, Jaeger being the default.
This may be good as an initial environment - as it would be possible for researchers to query the trace data from either Jaeger's query service, directly from the storage, or from a Kafka stream - depending upon how Jaeger is configured.
The downside is that this trace data would be in a Jaeger specific format.
As a future step we could provide additional instructions to configure OpenTelemetry collectors within this environment - this would enable research focused "processors" to be added to a custom OpenTelemetry Collector pipeline, and kafka stream based processing - but based on a "standard" (i.e. vendor neutral) trace data format.
Is the #1 stack sufficient for initial projects?