stream_partition in AddFields resolves to same value across all partitions
#77722
Replies: 1 comment
-
|
Hi Andreas Skielboe (@askielboe), thanks for the very detailed write-up — the reproduction matrix ( We're not able to confirm a root cause from the description alone, so we've escalated this to our team for investigation: airbytehq/oncall#12159. They'll follow up with any findings or follow-up questions there or back here. A few things that may be useful while the team takes a look: Possible workaround to unblock yourselfSince If your These are workarounds while we investigate, not a confirmation that the behavior you saw is incorrect. Information that would help us investigateIf you're able to share, the following would speed up the investigation on the linked oncall issue:
Need more help? Join the Airbyte Community Slack for peer support, or if you're a Cloud customer, open a support ticket referencing this discussion. |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
-
I have a declarative connector (Connector Builder, CDK 7.17.1) with three streams:
SubstreamPartitionRouterwith auth asparent. partition_field: auth_header+ request_option injects the Authorization header per tenant. Works correctly — each tenant's data is fetched with the right credentials.To pass
auth_headerfrom parent to child, I useAddFieldson the parent stream to stamp each record:The child then reads it via parent_key:
auth_header.Problem: The AddFields Jinja expression resolves to the same partition's value for every record across all partitions. All parent records end up with one tenant's auth header, so the child sends all requests with wrong credentials.
I've tried
stream_partition.<partition_field>,stream_partition.extra_fields['...'], renaming fields to avoid collisions, andconcurrency_level: 1— all produce the same result. Meanwhile,request_optionand URL templates correctly resolve per-partition.Is there a known issue with
stream_partitioninsideAddFieldsin the concurrent code path? Or a different way to propagate partition data onto output records?Beta Was this translation helpful? Give feedback.
All reactions