I am running scroll on multiple indexes which are built on timeseries data. I have an index for each day with a common prefix in indexname. Earlier I was giving index name wildcard to serve scroll but since it was using all indices, the performance was very slow.
So for optimization I am trying to implement scroll per index starting from latest index to oldest index. Now I am stuck at a place.
I fetch 10 records per scroll. When I am changing the index when all records of current index are consumed, I don't want to fetch 10 records from new index but just the remaining to make total as 10 records. From next scroll call I would want to fetch 10 records.
I think scroll doesn't provide changing the scroll batch size. Is there any way to handle this?