As refguide states, hbase.client.scanner.caching works with hbase.client.scanner.max.result.size to try and use the network efficiently.
Make sure the release you use is 1.1.0+ which had important bug fixes w.r.t. max result size. On Wed, Jan 11, 2017 at 9:46 AM, Josh Elser <els...@apache.org> wrote: > Behind the scenes, the ClientScanner is buffering results from the > previous RPC. Ignoring multiple RegionServers for now, the caching value > denotes the number of records that were fetched by the ClientScanner in an > RPC. When the buffered results are consumed by your client, a new RPC will > be made to fetch another 100 rows. > > Thus, larger scanner caching values reduces the number of RPCs the > ClietnScanner object will make; however, it increases the length of the RPC > itself (as it takes longer to aggregate a larger number of rows to return). > > Rajeshkumar J wrote: > >> I have hbase.client.scanner.caching as 100. I am scanning a table. For >> instance say we have 500 rows matching for the current scan. When I give >> the statement ResultScanner.next() what will happen? whether it will >> return >> 100 rows for each next operation or something else >> >> >> Thanks >> >>
