Incorrect members being shown in cluster in hac

[former_member736085]

Hi,

We have two environments prod , preprod . When prod db is refreshed on preprod env. , preprod env. Hac starts showing prod +preprod nodes in cluster monitoring console on hac.

This should not occur. Any pointers to where the problem can be. We have auto discovery turned on.

I can see cluster island id is same in both the environments. Any pointers on how to update cluster islandid?

[Eddy_]

We had the same problem after database replication from production to testing system. We have found out that the root cause is the Cluster Island ID which is derived from the schema of the master tenant. After DB replication, this ID was the same on TEST and PROD. Thus, the system believed that all nodes belong to the same cluster.

Changing the channel name only is not enough and will end up in warnings like:

WARN [jgroups-7861,hybris-broadcast,hybrisnode-2] [UDP] JGRP000012: Nachricht von Cluster hybris-broadcast-test wurde verworfen (unser Cluster ist hybris-broadcast). Der Sender war 36f19a1b-89ce-3fb8-0468-e5e75bf9a5b4 (7 identische Nachrichten erhalten von 36f19a1b-89ce-3fb8-0468-e5e75bf9a5b4 in den letzten 67621 ms)

Additionally you have to use different mcast ports as well. Property for that is:

# Default: 45588
cluster.broadcast.method.jgroups.udp.mcast_port=45589

To be safe, we changed the mcast address as well:
# Default: 224.0.0.1
cluster.broadcast.method.jgroups.udp.mcast_address=228.0.0.1

PS: Config to change the channel name:
# Default: hybris-broadcast
cluster.broadcast.method.jgroups.channel.name=hybris-broadcast-test


Kind regards
Eddy