Hello, I'm currently working on an upgrade of a Mysql database that is used for Druid metadata. We use AWS Aurora serverless Mysql 5.7 and will upgrade to mysql 8. Our connector will support this new version, so no issue there. However among all attention points brought by AWS, one of them bothers me. Any upgrade path could lead to some downtime (from less than 1 mn with blue/green deployment to several for in-place/snapshot-restore.) Since I don't want to have any interruption on my Druid service (or at least limit the blast radius of this operation) , I started to dig on that subject and did not found that much information. We're targeting the blue/green deployment that should have less impact on our users, but still. If the metadata database is not reachable for 1 minute, what could happen to Druid (we'll be able to stop ingestion during this timeframe) ? Will happen anything that could not be recovered when the database comes back online ?

Hi Giom, Using Imply Hybrid's deploymeny model in AWS, the RDS blue/green update there triggered only momentary disconnect/reconnects, and during this time there was no noticable disruption to cluster operations, either on the ingestion or querying side. I can't comment on orchestrating this manually using OS Druid though.