Databricks networking deep dive part3

[@RAFMAF-h8y]

Thank you so much for these videos. We are trying to implement a secure Databricks architecture where everything is locked down, with no inbound or outbound internet connection from Azure at all. Everything is forced through the hub firewall in Azure to the on-prem firewall. We noticed that when we create private DNS zones, some A records are overridden because we have only one private DNS zone for both (front-end-private-endpoint and back-end-private-endpoint) since they have the same name. We have a centralized private DNS zone solution created once and shared by all entities in the shared resource hub. Have you encountered this issue in a similar architecture?

[@Stateoftheheart]

Really helpful! Thanks a million for breaking this down!

[@ankitdagadu510]

I really appreciate your work, the series is really helpful. Databricks platform/Networking related videos are rare on YT thanks for putting efforts and helping the community. If possible for you to create a video on External Metastore configure with SQL/Postgres, Trino/Presto on AKS/VM

[@nikofeng8447]

Love it man, really appreciate your explanation, helps me a lot

[@prashantmait07]

awesome work brother

[@coding3438]

Thanks for the work you do brother

[@MAMIR_BLOG]

Thanks for such detail information, one question, if the user machine is not a cloud VM (under MainDataVNET), but it is a laptop connecting from corporate network through VPN/P2P/Express Route, does the FrontEnd Privatelink scenario (between user and data plane) still needed?

[@Est0qu3]

Super useful. But I have one question. You mention about one workspace just for browser_authentication. I do not see where you were creating it. Could you explain in more details please?

[@harryjain31]

wow, it is so much clear now.

[@MarcValk]

Thank you Mohammed for explaining all of this. One question: In your video you created 2 separate VNETS, 1 specific for databricks and one for the Client VM’s. Would it also work to create one VNET consisting of the 2 Databricks subnets, 1 subnet for the private endpoints and some more subnets like the client workstations. So everything in one VNET, or is that not recommended?