-
Bug
-
Resolution: Done
-
Minor
-
None
-
None
When running JGroups using IPv6 bind and multicast addresses, the following messages appear on the console:
> 20:21:36,565 WARN [UDP] failed to join /224.0.75.75:7500 on vif0.0:
> java.net.SocketException: No such device
> 20:21:36,566 WARN [UDP] failed to join /224.0.75.75:7500 on peth0:
> java.net.SocketException: No such device
> 20:21:36,566 WARN [UDP] failed to join /224.0.75.75:7500 on xenbr0:
> java.net.SocketException: No such device
The multicast address 224.0.75.75 is used by the JGroups diagnostics handler (in TP.java) to listen for incoming probe requests. Probe requests are issued from clients using Probe.java to find out which JGroups members are using UDP in a given network.
These messages expose two issues:
JGroups is trying to join a multicast group on certain Xen virtual interfaces (vif0, peth0, xenbr0) and this is failing
(ii) JGroups is using a fixed IPv4 multicast address, despite the fact that IPv6 addresses are being used as bind_addresses and multicast addreses.
Something similar to the first issue was addressed in https://jira.jboss.org/jira/browse/JGRP-167..
As for the second issue, shouldn't JGroups use default IP addresses which correspond to the IP mode the JGroups instance has been started in (i.e. if JGroups instance bound to IPv4 bind address, use IPv4 diagnostics multicast address; if JGroups bound to IPv6 global address, use IPv6 global diagnostics multicast address, etc)
This may very likely cause problems if JGroups is run in an IPv6 only environment, where IPv4 socket use is disallowed.