Waseem Taj <[EMAIL PROTECTED]> writes: > > W are testing integration of Sequioa with our existing monitoring > tool. The monitoring tool connects to the sequoia controller port to > show there's a process listening there. If the connect(2) call > works, it closes the connection and reports all OK, otherwise it > flags an error (we will use this monitoring in production to ensure > that we are alerted via email as soon as sequoia goes down > unexpectedly). However it seems that whenever the monitoring tool > attempts a connection, sequoia is throwing an error. This does not > seem to have any impact as the monitoring tools still reports all > OK, but it is filling the log files with the following error: >
>> 2006-12-07 14:49:53,765 ERROR controller.core.Controller I/O error > >> during driver connection (java.io.EOFException) > > > Can this be fixed? > Yes. As you can guess, the driver is also monitoring the controllers so you can just fake a driver and "ping" the controller just like drivers do. Just send(int org.continuent.sequoia.common.protocol.Commands.Ping = -1) to appease the controller before closing the TCP connection. You have to use java endianness for sending this 32bits int. (Well,... since this is "-1" it probably does not matter :-) This is the old TCP-based ping. To bypass TCP timeout issues the latest sequoia code is now moving to a new UDP ping instead. Check SEQUOIA-506 and the source code linked from there for more details. The old TCP ping will remain for compability reasons, it will NOT disappear. So you can opt either for TCP or UDP, your choice. The TCP ping is much easier to (re-)implement but fast/timely detection is not possible. _______________________________________________ Sequoia mailing list [email protected] https://forge.continuent.org/mailman/listinfo/sequoia
