This message was deleted.

Hi, could you please share why you want to compact a specific set of segments?

we have late data, rest of my segments are in good numbers , so i don't want them to be processed again, only segments with less size, One more info i am using rollup as well.

Ah, I doubt that it would work If you compact only a subset of segments, only this subset's data is likely to be present in the next version, while the remaining data would be overshadowed

ok it would not be available for query ? should i try reindexing in that case , will that help ?

Reindexing and compaction would have to process the entire data which is probably what you want to avoid, right?

one more case is , due to segment are big in size i also get OOM in the compaction task for whole day, so thought of only compaction small size segments, seeting inputsegmentsizebytes less makes whole interval to be ignored.

I think it's necessary to compact all the segments of one period, otherwise you will hide the other data from the other segments for the OOM, you can tune maxRowsInMemory and maxBytesInMemory, try setting a very low limit, like 20mb if it still getting OOM (heap) then, there's something strange, it's likely that you need to tune also max direct memory to something big (eg: 1.5x the number of processing buffer * number of threads * mumber of slots workers of the MM)

Thanks @Renato Santos option 2 worked, i had kept it low, but now made it more lower and it worked.