functioner commented on pull request #2727:
URL: https://github.com/apache/hadoop/pull/2727#issuecomment-787681075
> @functioner see TestRPC#testClientRpcTimeout
> Maybe It is expected that Send ping all the time when
_ipc.client.rpc-timeout.ms_ less than 0.
> @iwasakims Do you have any ideas? Thanks.
@ferhui See TestRPC.java line 1419-1426 (within testClientRpcTimeout)
```
try {
// should not time out because effective rpc-timeout is
// multiple of ping interval: 1600 (= 800 * (1000 / 800 + 1))
proxy.sleep(null, newSleepRequest(1300));
} catch (ServiceException e) {
LOG.info("got unexpected exception.", e);
fail("RPC should not time out.");
}
```
According to the comment there, I think the correct behavior is to assign
the effective rpc-timeout with multiple of ping interval.
In the current implementation (e.g., the constructor of
hadoop.ipc.Client$Connection), rpc-timeout is set as a single ping interval,
rather than multiple of ping interval.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
