kfaraz commented on issue #18008:
URL: https://github.com/apache/druid/issues/18008#issuecomment-2940942104

   > I'm slightly against supervisor UUIDs since that might make things less 
readable on the console (users would have to look at the spec to get an idea of 
what topic, etc. it's pulling from).
   
   @jtuglu-netflix , that is easily remedied since the web-console is powered 
by the `sys.supervisors` table which itself reads
   from an Overlord API. So, it would be fairly straightforward to add a 
datasource column to web-console.
   We would need to add this column anyway regardless of the approach we take 
since supervisor IDs can now deviate from
   the datasource name.
   
   > My thought is we keep the same supervisor ID scheme and just validate as 
we normally do via the id uniqueness constraint in druid_supervisors table.
   
   Do you mean something like supervisor 1 with `id = user-given-id-1` and 
supervisor 2 with `id = user-given-id-2`,
   both can write to the same datasource? 
   That would work too. In fact, that is what I had in mind when I mentioned 
removing the 1:1 relationship between
   datasource and supervisor in [this older 
comment](https://github.com/apache/druid/issues/18008#issuecomment-2915281098).
   
   Optional: We might also want to consider adding some more validations in the 
same vein as #17955
   so that users don't accidentally update a supervisor when they actually 
intended to create a new
   supervisor to write to the same datasource but read from a different 
topic/cluster.
   
   > I agree that druid_dataSource would need a supervisor_id column. As for 
APIs, I feel like keeping them the same for now is fine. The only new addition 
might be getting the full list of supervisors for a given datasource, which can 
be done by pulling the list of supervisors from druid_supervisors and filtering 
on datasource outside the query.
   
   Sounds good. This new API need not read from metadata store though, the 
`SupervisorManager` should
   already have all the necessary info cached in memory.
   Also, if we do the modifications with the `sys.supervisors` table suggested 
above, we probably wouldn't need this API.
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to