Jaeger - Distributed Tracing Platform
What is difference between jaeger collector and jaeger agent? #1488
-
|
Why do you directly send trace spans to jaeger collector from jaeger client? What does jaeger agent do? |
Beta Was this translation helpful? Give feedback.
Replies: 4 comments 1 reply
-
|
Agents are local to the host, so the client libraries do not need to do any service discovery to find them. Agents mostly just forward traffic to collectors, but they can also enrich spans with extra metadata (using agent tags).
|
Beta Was this translation helpful? Give feedback.
-
|
@yurishkuro Just an additional question, please. So why do we need to specify the collector endpoint in each client? I mean if the traffic is client -> agent -> collector, why should a client be aware of the collector endpoint? |
Beta Was this translation helpful? Give feedback.
-
|
you don't need to specify collector endpoint. If you do, the client will talk directly to collector, bypassing the agent. |
Beta Was this translation helpful? Give feedback.
-
|
Would Jaeger be able to send traces to a third party collector in zipkin V2 format? I have a use-case like this where I am trying to use Jaeger(in Python) to generate traces and transport the traces using HTTP to a third party trace collector service. Thanks! |
Beta Was this translation helpful? Give feedback.
-
|
FYI this is also documented in the FAQ on the documentation website |
Beta Was this translation helpful? Give feedback.
Agents are local to the host, so the client libraries do not need to do any service discovery to find them. Agents mostly just forward traffic to collectors, but they can also enrich spans with extra metadata (using agent tags).
For legacy reasons, agents implement HTTP endpoints that clients use to poll for sampling strategies and other configurations. The same HTTP endpoints are not mounted on the collectors.(this has been fixed)