Closed marandaneto closed 2 years ago
SDK's don't implement the dynamic_sampling_context_frozen logic, do we need it?
That's also about to change for the Java SDK with this PR https://github.com/getsentry/sentry-java/pull/2226
Do we just make setTraceSampling(boolean)
a NOOP and have isTraceSampling()
always return true?
Do we just make
setTraceSampling(boolean)
a NOOP and haveisTraceSampling()
always return true?
How about we just flip the default to true
. Possibly add an Obsolete annotation to mark for removal. This past would be best done when we have the new option tracePropagationTragets
to point users to.
@adinauer @bruno-garcia that was already described here https://github.com/getsentry/sentry-java/issues/2244
Let's move the conversation in there to not polute this meta issue, I agree with the flipping + deprecating, adding tracePropagationTragets
is also very trivial, pretty much a renaming of tracingOrigins
, that would also deprecate and point to the new one.
How about we just flip the default to true. Possibly add an Obsolete annotation to mark for removal. This past would be best done when we have the new option tracePropagationTragets to point users to.
@bruno-garcia that would also mean, that once the flag is removed, each "captureTransaction" will send the traceContext (dynamic sampling context) to sentry and there is no way for the user to disable this behavior. Can you confirm that this is the behavior we want?
tracePropagationTragets
That's the goal of tracePropagationTragets
, you can filter out specific URLs or everything.
@philipphofmann @brustolin is this being used?
is this the equivalent for isTraceSampling
on Java that should be replaced?
No, we dont have this flag anymore. We need to remove this from test.
No, we dont have this flag anymore. We need to remove this from test.
Please do it to avoid further confusion.
@brustolin, please create an issue for that on the Cocoa repo.
@brustolin or @marandaneto, do we already have an issue for that on the Cocoa repo? If yes, is it done?
user.segment
isn't unified. iOS reads fromuser.data
and Java fromuser.other
. https://github.com/getsentry/team-mobile/issues/37 but the right behavior is adding a top-level property calledsegment
. https://github.com/getsentry/sentry-javascript/blob/beb3c973c0f10c538c8369983fc08c5bd031bfde/packages/types/src/user.ts#L8 Java issuetracePropagationTargets
isn't unified, but we'd need to offer at least one option. Java reads fromtracingOrigins
and iOS does not have it at all. https://github.com/getsentry/team-mobile/issues/39baggage
behavior isn't unified. Java always overwrites thebaggage
header which is wrong per spec, iOS does not. https://develop.sentry.dev/sdk/performance/dynamic-sampling-context/#baggage Java has an open draft that addresses that.traceSampling
SDK option. Java is guarded by this option, iOS does not have it at all, not sure if such a flag should exist. Apparently, it's only gated bytracePropagationTargets
, we don't need suchtraceSampling
flag. Java issueSDK's don't implement the
dynamic_sampling_context_frozen
logic, do we need it? https://develop.sentry.dev/sdk/performance/dynamic-sampling-context/#unified-propagation-mechanism https://develop.sentry.dev/sdk/performance/dynamic-sampling-context/#freezing-dynamic-sampling-context Java has an open draft that addresses that. This is experimental and we should not do anything for now, even less relevant for Mobile.How do we track the adoption of DS on SDKs? do we need to?
Remove
isTraceSampling
option so we can propagate baggage by defaultRelated issues