Question on Upstream lineage (UpstreamLineage): I...
# ingestion
w
Question on Upstream lineage (UpstreamLineage): In the bootstrap_mce.json, I see that UpstreamLineage is defined at DatasetSnapshot level.
Copy code
{
  "com.linkedin.pegasus2avro.dataset.UpstreamLineage": {
    "upstreams": [
      {
        "auditStamp": {
          "time": 1581407189000,
          "actor": "urn:li:corpuser:jdoe",
          "impersonator": null
        },
        "dataset": "urn:li:dataset:(urn:li:dataPlatform:kafka,SampleKafkaDataset,PROD)",
        "type": "TRANSFORMED"
      }
    ]
  }
}
Shouldn't there be a feature to track lineage at field level? @green-football-43791 @big-carpet-38439
g
field level lineage is on our roadmap: https://datahubproject.io/docs/roadmap
looks like we need to update those timelines a bit, but this is something on our near term radar that we aim to start tackling in the next few weeks.
😐 1
1
do you have specific column level lineage use cases in mind?
are there specific sources or goals you have in mind for column level lineage?
w
Yes, definitely. I'm working on EDD and Metadata lineage at our org (embibe.com)
I started with Neo4J, but felt that the metadata Type system is already thought after in DataHub, so proceeded with the metadata modeling.
g
gotcha- we'd love your input during the design process.
w
Field level lineage is very important to us and I think for most of the org. 1. E.g. as a product owner I want to see from the dashboard which all fields (standard and derived) are contributing to my dashboard and from where are they originated.
g
we have a channel #column-level-lineage for discussing this feature specifically
so feel free to join and give your input there 🙂 it would be great to hear your thoughts.
w
alright, I'd follow
Looks like I've to fall back to Neo4J and use my own custom relationships between nodes to build a metadata model for my org.
h
Hi @witty-keyboard-20400, the API for column-level lineage will be coming soon, followed by the v1 UI a little later. We recommend using the APIs rather than the custom relation approach if possible.
👀 1