On Wed, Jul 24, 2013 at 9:55 PM, Henry Hung <[email protected]> wrote:
> Hi All, > > I'm a HBase newbie. Today I start to do stress test using 4 java process > to load large amount of rows (1 minute = 18,000 rows). After a while, one > of my region server log got this kind of WARN message: > > WARN org.apache.hadoop.ipc.HBaseServer: IPC Server listener on 60020: > readAndProcess threw exception java.io.IOException: Connection reset by > peer. Count of bytes read: 0 > > Question: > > 1. What is the meaning of this WARN message? > > This is pretty good on the 'Connection reset by peer': http://stackoverflow.com/questions/1434451/what-does-connection-reset-by-peer-mean The message is from our internal IPC/RPC listener thread. It is trying to read a query from the client but the client gave up on us before we could read anything (0 bytes). > 2. I haven't check the data integrity, but is there any chance of > data loss when encounter this WARN message? > > Client side we'll retry up to a configurable amount of times. If we do not succeed, if the client fails to write, you will see an exception in your application complaining that this is the case. > Additional info: > For this stress test, I have 10 servers (with each server having region > server and data node). > Only one server has that WARN message. > Would suggest you add data validation to assuage any concerns you have about data loss. St.Ack
