This message was deleted.
# generals
Slackbot
07/19/2023, 3:21 PMThis message was deleted.
d
Didip Kerabat
07/19/2023, 6:35 PMWe have a partial answer to that.
For very important customers, we give them a dedicated historical replication tier and simply deploy special historical StatefulSet just for them.
But our broker tiers are still shared. Which is fine for now… Broker is not the bottleneck.
s
Sergio Ferragut
07/19/2023, 6:38 PMHi Siva, seems like you are referring to three separate functions:
• datasource level security - where you use access control to restrict usage of data sources by user/role
• datasource resource usage isolation (this is what Didip refers to) segregate datasources by historical tier so it is dedicated to a tenant that has access to that data source)
• efficient data filtering on a single data source that has data for multiple tenants where you use secondary partitioning (CLUSTERED BY in SQL based ingestion) to only access segments that are relevant to the tenant at query time.
Sergio Ferragut
07/19/2023, 6:42 PMSome docs that can help:
• https://druid.apache.org/docs/latest/operations/security-user-auth.html#authentication-and-authorization-model - control access by datasource
• https://druid.apache.org/docs/latest/operations/mixed-workloads.html#historical-tiering - setup tiers
• https://druid.apache.org/docs/latest/operations/rule-configuration.html#set-retention-rules - setup datasource to tier configuration
• https://druid.apache.org/docs/latest/multi-stage-query/concepts.html#clustering - use secondary partitioning on tenant_id
s
Siva
07/20/2023, 6:32 AMThanks Didip and Sergio for response and giving me clarity.