How incremental repair works if compactions happen before next incremental repair

Question

I want to perform repairs,when going through various options to reduce the time of repairs i came across these two options incremental repairs and performing parallel repairs on partition range.

1.) If i perform incremental repairs, it follows anti-compaction and separates Repaired and Unrepaired SStables. Next time before i perform incremental repair if the repaired and unrepaired sstables are compacted together then how incremental repair will be performed on this SStable will it perform repair on whole sstable or is there any way it will differentiate ??

2.) If i go with parallel repairs on partition range will it have a significant impact on the cluster ??

Answer 1

My first question before answering this would be "What is your use-case, what compaction strategy are you using and what are you targeting?". Moving to incremental repairs when using leveled compaction strategy and aiming for low latencies is a good decision. Having that in mind, I'll try to answer your questions.

1.) Repaired parts are compacted and un-repaired parts are size tiered. So if you are using leveled compaction repaired parts are leveled and un-repaired are size tiered.

2.) Parallel repair is usually used in operational downtime or when you want to finish repair quickly. They are resource hungry so dont run them scheduled in production.

All in all it depends on your use-case and what you are aiming for.