jgroups-channel for messaging communication

[developer251]

Hi all!

We are planning to use a static list of servers for messaging, inherited from JGroups. I have seen that you have to provide the jgroups stack and jgroups-channel attribute in the broadcast and discovery groups. What we want to achieve is to have two master and two backup nodes in the cluster. Is it correct the following configuration:

Server1

<jgroups-channel>live</jgroups-channel>

Server2

<jgroups-channel>backup</jgroups-channel>

Server3

<jgroups-channel>live</jgroups-channel>

Server4

<jgroups-channel>backup</jgroups-channel>

Is it a good idea, or (since the number of servers is static), if Server1 fails for example, there will be a single point of failure for the cluster ?

Should we rather use for all servers:

<jgroups-channel>hornetqcluster</jgroups-channel>

Thanks

[jbertram]

I recommend using normal static clustering (demonstrated in the "Clustered Static Discovery" example shipped with HornetQ) instead of JGroups.  All servers in the cluster can have the same list of static servers.  If you're using a shared-store configuration then your live and backup will simply share the same journal, but if you're using replication then you can pair a specific live and backup together using <backup-group-name> as discussed in the HA documentation.

[developer251]

Hi Justin,

thanks for the reply. By adding the static list of connectors results the following obscure error when starting the server:

Caused by: com.ctc.wstx.exc.WstxParsingException: Received non-all-whitespace CHRACTERS or CDATA event in nextTag().

at [row,col {unknown-source}]: [340,5]

I've attached the configuration file I have used which seems XML compliant anyway. Do you know if it's a bug ? (I've tested on releases EAP 6.2.2 and 6.3.0)

Thanks

[jbertram]

All your <cluster-connection> elements should be wrapped in a <cluster-connections> (notice the s character).  Take a look at docs/schema/jboss-as-messaging_1_4.xsd for more details on the schema.

[developer251]

Thanks for your reply Justin.

I've wrapped the list of cluster-connection elements, as indicated in 19.4.4. Switch UDP to TCP for HornetQ Clustering

However the problem remains. I'm receiving the same error. Is it a bug ?

Thanks

[jbertram]

There are some bizarre characters in your configuration file that are causing parsing to fail.  Read the error message and find the row/column coordinates that it reports and that is where a non-parsable character exists.  Once I cleared those out the server started just fine.

[f_marchioni]

Max I suspect you could be using JDK 1.8 which is not a certified platform on EAP 6.2.2. If you cannot find a solution, you might look at this article which shows how I managed to run it on EAP6 and WildFly