This message was deleted.

@Didip Kerabat Based on past slack history , it seems you may have valuable inputs to add on these two question . Apologies for tagging you directly but would love hear your thought on these two topics [data migration] and [ Druid Rolling upgrade | helm | k8 ]

1. S3 bucket migration can be done by simply copying the files. 2. But for metadata storage, you have to do extra work: you need to dump the segments table, rewrite the path to new S3, and then re-insert the metadata into new metadb. 3. I don’t recommend the new cluster reaching into old cluster to download historical files. They should be kept completely separate.

Thanks Didip . I understand the recommended steps . I read long back in one of the threads where you were reading from the same deep storage and meta db (read cluster / write cluster kind of setup which helped you achieve proper upgrade without down time ) . In my case , what are the problems you think can happen ? Essentially, the new cluster is reading from two S3 locations but writing into its own deep storage and not touching the other metadb at all. On the metadata part it just has metadata data copied over for a non overlapping interval which points to old cluster S3 path and hence is able to load without any meta data change .