Hello everyone, I'm wondering why <https://github....
# feature-requests
t
Hello everyone, I'm wondering why https://github.com/linkedin/datahub/blob/master/li-utils/src/main/pegasus/com/linkedin/common/FabricType.pdl has only pre-defined set of values. We would like to split our hierarchy of objects in UI based on GEO locations like NA, EMEA, AP etc., but looks like the only option is to have {FabricType}/{data source type}/{database name} and there is no flexibility for custom value on any of these layers. Am I missing something?
a
+1, they do have this as a feature request here: https://github.com/linkedin/datahub/issues/3001
b
Hi there!
Yeah so a couple things: 1. We are currently working to remove the tight restriction on this enum by converting it to a string. We should have a PR up for that soon. The purpose of the string is mainly to differentiate different "logical environments" or clusters of the dataset. 2. We are also working on an initiative to build out the concept of "Containers", which will allow you to bucket groups of related resources into common categories and view them together. I think this feature is really what would help address your use case. Does that sound correct?
a
We are also working on an initiative to build out the concept of "Containers", which will allow you to bucket groups of related resources into common categories and view them together. I think this feature is really what would help address your use case. Does that sound correct?
Yeah containers will be helpful for sure for discovery and potentially other things. But similar to Mikalai, the bigger "problem" is we have multiple Snowflake instances and MySQL clusters so the current datasetUrn isn't able to uniquely identify datasets. e.g. this URN clashes in different Snowflake instances
urn:li:dataPlatform:snowflake,db.schema.table,PROD
. If we can somehow include the instance/cluster information then that would work for Snowflake + MySQL.
b
Yes - we are very aware of this modeling problem... we are still figuring out the most ideal way to solve it.. One approach as you mentioned is to add a cluster / data platform instance id into the URN itself... More details to come soon on this... cc @mammoth-bear-12532 @little-megabyte-1074
t
Hi @big-carpet-38439 @mammoth-bear-12532 your planned changes sounds great! One interesting thing I see is Transformers for recipes. According to documentation they should provide ability to create aliases like "/PLATFORM/marketing_db/DATASET_PARTS". Don't you see it as alternative? In fact it doesn't work for me for some reason. I see folder structure in UI, but not dataset itself. See screenshot
b
Hmm this is very interesting.. What is the full browse path you are creating when populating the BrowsePathAspect??
l
Hi folks! I’ve migrated this over to our new feature request portal; we are targeting supporting multiple platform instances before the end of the year! I’d encourage you to subscribe to the post for updates as we progress https://feature-requests.datahubproject.io/b/User-Experience/p/support-multiple-platfrom-instances