You are viewing a plain text version of this content. The canonical link for it is here.

Posted to dev@hbase.apache.org by Ted Yu <yu...@gmail.com> on 2011/10/26 00:05:06 UTC

load testing results for HBASE-4508

>From Shrijeet:

1. Rolling restart with 0.90 trunk + patch jar : PASSED
2. Verify that 0.90 trunk jar leaks connection with  HTable pool size set to
10 , hbase.zookeeper.property.maxClientCnxns set to 20 and launching 10
client threads : PASSED
3. Verify that 0.90 trunk jar + patch does *not* leak connection with  HTable
pool size set to 10 , hbase.zookeeper.property.maxClientCnxns set to 20 and
launching 10 client threads : PASSED
4. Verify that 0.90 trunk jar + patch still holds up when HTable pool
size set to 10 , hbase.zookeeper.property.
maxClientCnxns set to 5 and launching 100 client threads : PASSED

The established connections in my final test was 2. This I verified using
both netstat and zookeeper dump in UI.

One caveat: remember to set hbase.connection.per.config false in order to
use connection sharing.

5. hbase.connection.per.config=true , max connections to 300 and bumping
ulimit from 1024 to 2048, client threads 100 : PASSED

Kudo to Shrijeet.