[
https://issues.jboss.org/browse/JGRP-2343?page=com.atlassian.jira.plugin....
]
Bela Ban commented on JGRP-2343:
--------------------------------
I don't think we need to look at _all_ addresses; it is probably sufficient to
configure {{bind_addr}} in the transport (UDP, TCP) and then determine the IP type for the
rest of the addresses to be resolved (listed above) from it.
E.g. if {{bind_addr}} is IPv6, {{UDP.mcast_addr}} will be the IPv6 default (if not set),
or will be mapped to an IPv6-mapped IPv4 address (if set to an IPv4 address).
Change in IPv4 versus IPv6 picking algorithm
--------------------------------------------
Key: JGRP-2343
URL:
https://issues.jboss.org/browse/JGRP-2343
Project: JGroups
Issue Type: Task
Reporter: Bela Ban
Assignee: Bela Ban
Priority: Major
Fix For: 4.1.0
When
https://issues.jboss.org/browse/JGRP-2305 was implemented, IPv4 was favored unless
{{java.net.preferIPv6Addresses}} was set.
However, this makes the following case fail: no {{java.net.preferXXX}} option is set and
the configuration contains one or more IPv6 addresses, e.g. {{UDP
bind_addr="::1"}}.
This will fail, as the new algorithm sets the stack type as IPv4 (it works when
{{java.net.preferIPv6Addresses}} is set).
Change:
* Look at all IP addresses in a configuration
* If all addresses are IPv6 and the stack type is IPv4 (and IPv6 is available) ->
switch to IPv6 as stack type
* If all addresses are IPv4 and the stack type is IPv6 (and IPv4 is available) ->
switch to IPv4 stack type
--
This message was sent by Atlassian Jira
(v7.12.1#712002)