search

getsentry / sentry-docs

Sentry's documentation (and tools to build it)
https://docs.sentry.io
Other
332 stars 1.45k forks source link

Laravel note on performance overhead potentially misleading #11337

Open kosty-sentry-dw-eval opened 1 month ago

kosty-sentry-dw-eval commented 1 month ago

SDK

PHP SDK

Description

https://docs.sentry.io/platforms/php/guides/laravel/#local-development-and-testing Mentions negative effect of Sentry SDK on performance. I believe it may be a bit misleading

  1. This may lead customers to generalize that to Sentry overall, however this is something specific to only PHP platform that doesn't allow sending events in the background. Overhead on all other platforms is negligible as background threads (or equivalent) are used.
  2. AFAIK even for PHP (in production environment) it can be effectively mitigated by running a local relay in the same datacenter that reduces latency. We only mention that in the deep fine print of [Performance Monitoring] Performance Overhead and only in the context of AWS Lambda (I believe it should apply universally). So customer may conclude overhead locally => overhead in production and not be aware of mitigation. And this Laravel passage in question doesn't link to that doc.

Context: this section seems to be picked up a lot by RAG as it matches certain use cases well and other platforms don't have an equivalent doc. It's on RAG that it wasn't able to tell it's SDK-specific, but maybe we can meet it halfway? 🙂

Suggested Solution

add something along the lines of "because PHP unlike other platforms doesn't have background threads"? (@cleptric correct me if this is wrong please)

getsantry[bot] commented 1 month ago

Assigning to @getsentry/support for routing ⏲️

cleptric commented 3 weeks ago

We got PR open at #11533. I just want to make sure that we fully understand how our SDKs work. Just because an SDK uses an async transport, it does not mean that the overhead magically disappears.

realkosty commented 3 weeks ago

I just want to make sure that we fully understand how our SDKs work. Just because an SDK uses an async transport, it does not mean that the overhead magically disappears.

yes, also 'overhead' is not very precise, the answer is different depending on what you mean by it :)

You most likely don't want errors to be sent to Sentry when you are developing or running tests

curious why is SDK overhead a bigger problem when developing or testing rather than production? I understand you want quick local iterations but still there's something off in that logic...