chore(integrations): Add SLOs to finish_pipeline

[Christinarlong]

I dont think we have any SLOs for finish_pipeline despite being a major part of our integration installation pipeline

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

✅ All tests successful. No failed tests found.

Additional details and impacted files

@@            Coverage Diff             @@
##           master   #94586      +/-   ##
==========================================
- Coverage   87.95%   87.94%   -0.01%     
==========================================
  Files       10386    10393       +7     
  Lines      601771   602459     +688     
  Branches    23408    23408              
==========================================
+ Hits       529258   529809     +551     
- Misses      72057    72194     +137     
  Partials      456      456              

[cursor]

Bug: Permission Errors Masked as Success

Integration installation permission failures are incorrectly recorded as successful in SLO metrics. The finish_pipeline method returns early on permission errors without calling lifecycle.record_failure() or lifecycle.record_halt(). As the IntegrationPipelineViewEvent context manager defaults to assume_success=True, these failures are logged as successes, skewing reliability metrics. This is inconsistent with how other errors (e.g., IntegrationError) are handled.

src/sentry/integrations/pipeline.py#L139-L155

sentry/src/sentry/integrations/pipeline.py

Lines 139 to 155 in d85d837

	if (
	org_context
	and (not org_context.member or "org:integrations" not in org_context.member.scopes)
	and not superuser_has_permission(self.request, ["org:integrations"])
	):
	error_message = "You must be an organization owner, manager or admin to install this integration."
	logger.info(
	"build-integration.permission_error",
	extra={
	"error_message": error_message,
	"organization_id": self.organization.id if self.organization else None,
	"user_id": self.request.user.id,
	"provider_key": self.provider.key,
	},
	)
	return self.error(error_message)

Fix in Cursor

---

Bug: SLO Metrics Misreport Integration Failures

The IntegrationPipelineViewEvent context manager incorrectly records IntegrationProviderError cases as successful in SLO metrics. When an IntegrationProviderError occurs, the finish_pipeline method returns early via self.render_warning() without calling lifecycle.record_failure() or lifecycle.record_halt(). This causes the context manager, which defaults to assume_success=True, to log these errors as successes, skewing SLOs. This behavior is inconsistent with how IntegrationError is handled, which correctly records failures.

src/sentry/integrations/pipeline.py#L169-L180

sentry/src/sentry/integrations/pipeline.py

Lines 169 to 180 in d85d837

	return self.error(str(e))
	except IntegrationProviderError as e:
	self.get_logger().info(
	"build-integration.provider-error",
	extra={
	"error_message": str(e),
	"error_status": getattr(e, "code", None),
	"organization_id": self.organization.id if self.organization else None,
	"provider_key": self.provider.key,
	},
	)
	return self.render_warning(str(e))

Fix in Cursor

---

Was this report helpful? Give feedback by reacting with 👍 or 👎