GrapheneIntegration: Track operation name for tracing #2704
Conversation
keithhackbarth
changed the title
|
Hi @keithhackbarth, i would love to see this change! |
… query getsentry#2704 This commit adds a missing test for the graphene integration, so that we can be sure that a transaction with the corresponding query name is captured.
… query getsentry#2704 This commit adds a missing test for the graphene integration, so that we can be sure that a transaction with the corresponding query name is captured. --amend
|
Hey @keithhackbarth, i added a test and made a PR on your fork 😄 |
Feature/add graphene tests
… query getsentry#2704 This commit adds a missing test for the graphene integration, so that we can be sure that a transaction with the corresponding query name is captured. --amend
| with hub.start_transaction(op="graphql", name=kwargs.get("operation_name")): | ||
| result = old_graphql_sync(schema, source, *args, **kwargs) |
There was a problem hiding this comment.
We should be starting a span here rather than a transaction, since Graphene is likely to be called by a web server's response handler, which would already have a transaction automatically started for it (we integrate with the popular web frameworks, like Flask, Django, etc). Creating a transaction within a transaction leads to undefined behavior; they cannot be nested, so we need a span, not a transaction here.
Since spans do not have a name, the operation name should be stored in the span's description instead.
Lastly, the op should be one of the values listed under the GraphQL category in the "Usage" column of this table.
It is helpful to have the operation's name if trying to track the individual performance of GraphQL transactions using Sentry. This PR implements the approach suggested in this article, as part of the GrapheneIntegration()
Note: I looked and wasn't initially sure how to write a good test for this and also wasn't sure if this PR is even a good idea in the first place. So leaving a light version here for feedback.