getsentry / getsentry/sentry-python

Don't hold references to objects in telemetry pipeline

Open
#5,582 1 comment 0 reactions 1 assignee Claimed by @sl0thentr0py View on GitHub
Improvement Python Quality Improvement Span First
Dominant language
Python
Stars
2.2k
Forks
669
Avg merge
1d 1h
Merged PRs (30d)
213

Description

## Problem

A recurring problem in SDKs is that we hold references to userland objects throughout the event pipeline (scope apply, scrubbing) and sometimes that causes

* data corruption (modify user state)
* race conditions in concurrent flows

There is no real need for us to risk holding references since all we need is to collect data reliably and without locks and blocking user flow.

Some examples of this problem:

* django user lazy object instantiation: [https://github.com/getsentry/sentry-python/issues/5274]()
* go panic: [https://github.com/getsentry/sentry-go/pull/1214]()
* python scrubber modifying user objects: [https://github.com/getsentry/sentry-python/pull/3409]()

## Decision

* eventually once everything is on attributes API (including errors), we will make sure `set_attribute` doesn't hold direct references to userland state and reduce surface area of these races/data corruptions
* this can be through a lazy delegation wrapper object as in my [gist]()
* or just pre serialization in `set_attribute` itself


This is a placeholder task to decide and investigate the references that the span first pipeline holds and try to clean that up.

Contributor guide

Open the contributing guide

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.