Review sampling rate advice

[rhcarvalho]

While updating the docs in #1667 (comment), a conversation started about what advice we give for setting the sampling rate of different components, specially in the light of different load patterns.

The updated docs with the advice reads:

If you enable tracing in services with multiple entry points, we recommend choosing similar sampling rates, to avoid biasing your data. [...]

And the discussion is that even setting all sampling rates to the same value may not help with sample bias.

This is a screenshot of the advice prior to the update (red underline to emphasize the advice):

https://docs.sentry.io/performance/distributed-tracing/#sampling-transactions

[github-actions]

This issue has gone three weeks without activity. In another week, I will close it.

But! If you comment or otherwise update it, I will reset the clock, and if you label it Status: Accepted, I will leave it alone ... forever!

---

"A weed is but an unloved flower." ― Ella Wheeler Wilcox 🥀

[rhcarvalho]

The advice, despite a few edits, is still there at https://docs.sentry.io/product/performance/distributed-tracing/#consistency-between-traces.

@trillville I'd appreciate your help if you could have a look at that advice in the docs? My understanding is that, in that case, distribution bias is orthogonal to the choice of sample rates.

[trillville]

The advice, despite a few edits, is still there at https://docs.sentry.io/product/performance/distributed-tracing/#consistency-between-traces.

@trillville I'd appreciate your help if you could have a look at that advice in the docs? My understanding is that, in that case, distribution bias is orthogonal to the choice of sample rates.

I agree. Just to make it concrete with an example:

• 10k web app frontend transactions
• 10k backend transactions (from web app)
• 100k backend API transactions

If frontend sampling rate is 10% and backend is 10%, we end up with:

• 1k web app frontend transactions
• 1k backend transactions (from web app)
• 10k backend API transactions

If frontend sampling rate is 10% and backend is 50%:

• 1k web app frontend
• 500 web app (backend)
• 5k backend API transactions

In both cases, the ratio of backend transactions attributable to the web app to backend transactions attributable to the API is the same.

[trillville]

Let me know if I'm misunderstanding something! But if I'm not, definitely agree that different sampling rates wouldn't affect bias here.

[PeloWriter]

@lobsterkatie or @wmak - can either of you take a look and update the content here?