*hbase.client.ipc.pool.size* defines the number of conections(not threads) to each region server,.
I'm not very sure if it would work, but I'm sure you should not reuse a connection in a single jvm. Vitaliy Semochkin <[email protected]> 于2019年11月7日周四 下午7:59写道： > Thank you very much Yuhang Li! > > Do you think I also should increase > hbase.client.ipc.pool.size > if I have hundreds parallel threads executing Get? > > Regards, > Vitaliy > > On Thu, Nov 7, 2019 at 1:06 PM yuhang li <[email protected]> wrote: > > > > Yes,there is a thread pool in hbase connection. > > > > You can set the core pool size by *hbase.hconnection.threads.core*, and > the > > max pool size by *hbase.hconnection.threads.max*. > > > > Vitaliy Semochkin <[email protected]> 于2019年11月7日周四 上午6:48写道： > > > > > Hi all, > > > > > > I have many client fetching data from HBase > > > from a single jvm. When I reuse HBase connection the number of > > > requests per second that application can handle is relatively small, > > > however if I create a pool of connections and reuse it the number of > > > requests per second dramatically increases. > > > I suspect this is because HBase connection has a pool inside which is > > > relatively small. What parameter/parameters can be altered in order > > > to improve performance of single hbase connection? Are there any means > > > to increase maximum concurrent operations allowed? > > > > > > Regards, > > > Vitaliy > > > >
