Oracle internals, debugging and undocumented features
Issue with Oracle 220.127.116.11 new Redundant Interconnect
November 1, 2010Posted by on
MOS note 18.104.22.168 Patch Set – List of Bug Fixes by Problem Type [ID 1179583.1]
If you look for patch 9974223 you can see that patch is there:
under Undocumented Oracle Server
9969284 9969539 9971646 9972586 9973072 9973560 9974223 9974736 9975343 9975837
but Bug 9974223 :
Oracle Grid Infrastructure for a cluster needs multicast communication on 22.214.171.124 addresses
working properly between the cluster nodes.
“When first attempting to start an 126.96.36.199 cluster node, either during a fresh install or after an
upgrade attempt, it is possible that the node will be unable to join the cluster.
The Oracle Cluster Synchronization Services (OCSSD) will report a failure to join the
existing cluster members.This will be noted as an inability for the CSS to connect to the other active members of the cluster.
The problem occurs because the clusterware in 188.8.131.52 uses multicast messages to initiate communications on the 184.108.40.206 address.
If multicast on these addresses is not enabled for the private interfaces, then the clusterware will be unable to connect to it’s peer nodes.”
When I run mcast1 program ( MOS note 1212703.1 )on my nodes using 220.127.116.11 address it never locates new peer..
but when I use 18.104.22.168 it works..
Here is what Oracle says:Note ID 1054902.1
For 22.214.171.124 and above, private network, without patch 9974223, multicast address 126.96.36.199 should work; with the patch,
another multicast address 188.8.131.52 is supported. See Multicast section to verify.
Guess what.There is no patch 9974223 and multicast does not work for 184.108.40.206 address.
One of the solution I have found on internet
(if you are using Cisco switches) is to configure no ip igmp snooping on the switch.
And answer from MOS:
Oracle released Patch 9974223.It is available to download.