Issue Description
The docs give an example on how to adjust the sample rate depending on the context. However, it appears the example doesn't match what ultimately gets reported in the Sentry dashboard.
Reproduction Steps
Add this config from the docs and you'll see that the op is never http.server for any http requests, but instead atoms like :GET and :POST
config :sentry,
traces_sampler: fn sampling_context ->
case sampling_context.transaction_context.op do
IO.inspect(op, label: "Actual op")
"http.server" -> 0.1 # Sample 10% of HTTP requests
_ -> 0.05 # Sample 5% of other operations
end
end
Expected Behavior
Sample code should work, or docs should be updated to match what the library actually does.
Actual Behavior
100% of the traces reported by my Phoenix app show up with an op of either db or http.server on the Sentry dashboard. But the op field in the context passed to traces_sampler is always something more detailed. For example, if I issue a GET request to /health-check, this is what the context actually looks like when I inspect it.
%{
attributes: %{
"url.path": "/health-check",
"http.request.method": :GET,
"client.address": "127.0.0.1",
"network.peer.address": "127.0.0.1",
"network.peer.port": 62201,
"network.protocol.version": :"1.1",
"url.scheme": :http,
"user_agent.original": "curl/8.7.1",
"server.address": "localhost",
"server.port": 4001
},
name: :GET,
op: :GET,
trace_id: 68167592341524485616445320670865761234
}
However, in Sentry's dashboard, this shows up like this, with an op of http.server.
So if I try to filter on http.server like the example shows, it doesn't work because op doesn't become http.server until some point after the traces_sampler function runs.
The same issue for db ops. In the context, they will show up as my_app.repo.query but report as db on the dashboard.
Elixir Version
1.18.4
SDK Version
11.0.4
Integration and Its Version
No response
Sentry Config
No response
Issue Description
The docs give an example on how to adjust the sample rate depending on the context. However, it appears the example doesn't match what ultimately gets reported in the Sentry dashboard.
Reproduction Steps
Add this config from the docs and you'll see that the op is never
http.serverfor any http requests, but instead atoms like:GETand:POSTExpected Behavior
Sample code should work, or docs should be updated to match what the library actually does.
Actual Behavior
100% of the traces reported by my Phoenix app show up with an
opof eitherdborhttp.serveron the Sentry dashboard. But theopfield in the context passed totraces_sampleris always something more detailed. For example, if I issue a GET request to /health-check, this is what the context actually looks like when I inspect it.However, in Sentry's dashboard, this shows up like this, with an op of
http.server.So if I try to filter on
http.serverlike the example shows, it doesn't work becauseopdoesn't becomehttp.serveruntil some point after thetraces_samplerfunction runs.The same issue for
dbops. In the context, they will show up asmy_app.repo.querybut report asdbon the dashboard.Elixir Version
1.18.4
SDK Version
11.0.4
Integration and Its Version
No response
Sentry Config
No response