Hi,
we have a TTL of 3 years for few tables. We would like to alter the table to use TTL of 6 months. If we alter the table and delete data older than 6 months. It will create lot of tombstones. In this case the reads will be much slower . May I know how to safely delete data and evict tombstones without impacting performance?
Please suggest
And also if table migration is one option to look at?
If so how can I do without any downtime?
There isn't an out-of-the-box solution for this. You will need to write an app preferably with Spark that would:
If you are deleting whole partitions then tombstones will not be a problem. But if you are deleting rows within a partition then tombstones will be a problem if you are scanning through the deleted rows. There isn't any way to avoid that unless you redesign your data model and/or rewrite your app code. Cheers!
@Erick Ramirez Thanks. May I know the way to do any keyspace/table migrations without any downtime? Is it possible
Lets say If I want to restore the live prod table to other cluster ? where the application can use it as prod?
[Post converted to a comment since it's not an "answer"]
@bharadhwajt, a similar question was answered by Erick earlier and you could refer the procedure (it is for load, but you could refer to the blogs and examples for "unload" operation) to unload data from source keyspace/table and load it to destination keyspace/table provided the schema is same/compatible there using DSBulk
7 People are following this question.
