We are trying to upgrade from real-time nodes to supervisors in Prod environment and is there any way there can be a rolling update done without downtime and loss of data ?. Can you please provide some documentation on how that can be done ?
I think Druid doesn’t support it by itself, but you might be able to do it with an external stream controller which switches the stream receiver from realtime nodes to the supervisor.
Is it possible to stream data into an existing data source through supervisors ?
Yes, it’s possible, but only one supervisor can run per dataSource.
Thank you. How can we make sure that there is no duplication of data when we have real-time nodes and supervisors pointing at the same data-source running at same instance. My plan is to leave the real-time nodes running and start supervisor(wait for it to submit one task at-least) before stopping real time nodes which might avoid duplication of data.
I am currently seeing ingestion tasks appear even after real time services are stopped(has been more than 10 hours). Do you probably know why that would happen ?(https://groups.google.com/forum/#!topic/druid-user/CBPfM4KQ8Xg)