Checkpointing records migration progress so a long-running process can pause and resume without starting over. It is especially useful when a migration must be throttled, paused to scale the destination, or adapted to source-database limits. section 12.2.4
A migration can use checkpointing to:
- Record completed token ranges or other processed portions of the data set. section 12.2.4
- Pause before scaling the destination or reducing pressure on the source database. section 12.2.4
- Resume from the recorded position instead of repeating completed work. section 12.2.3
- Store progress in a table in either the source or destination database; the mechanism need not be sophisticated. section 12.2.4
Migrating data in ranges
A custom migrator divides ScyllaDB data into token ranges and stores each completed range in a checkpoint table. If the source must be throttled after several ranges, the process pauses; when it resumes, it continues with the remaining ranges rather than restarting the migration. section 12.2.4
