Cluster checks failing: Recreate tables

Hi @denniswi,

every migration story can be a bit different but I hope I can give you some pointers

Can I assume that the newly inserted records are all written in latest partition? In this case you can safely migrate all old partitions to a new table. This process can run in parallel while new records are being written to the old table.

Once all old records are backfilled you are left with only most recent partition to migrate. Here you can opt for a short downtime to migrate the last partition and switch over tables. If you want to further reduce the required downtime do this on a Monday morning, when you have a fresh partition with just a small number of records.

You also have a option for a zero-downtime table swap, but this will require some additional planning and work. For example you could deploy a code change which writes new records to both the new and old table. In parallel backfill the new table from the old table. Once you reach record paritiy you would have the chance to run some checks (completness, performance, etc) on the new table. If everything looks good do a switch over to the new table.

Hope this helps