symmetric cluster UPD other than on 231.7.7.7:9876

[Aaron Dutenhoefer <aa...@gmail.com>]

I'm using Artemis ActiveMQ v2.11.0 currently - have been using it since 1.x
days.  I have three environments:  Integration, QA, and Production.  In
integration, I've got a standalone broker instance.  In QA I'm running a
two-node symmetric cluster using UDP discovery.  I followed the docs and
that worked fine.  In Production I'm trying to set up a four-node symmetric
cluster.

When I set up my first production cluster node, I started it up and it
immediately joined my QA cluster.  This was expected since the config still
was using 231.7.7.7:9876 as the discovery address/port.  (My Int/Qa/Prod
VMs are in the same datacenter/VLAN so UDP traffic routes across them)   I
then reconfigured the settings using another UDP address - Only to find the
other nodes configured using the new multicast/UDP address:port never
connect or discover each other.

I've double-and-triple-checked my configs and even tried using the examples
and updating my working QA cluster to use another multicast addr (like
231.7.7.8 for instance).  Any other address causes the cluster to lose
sight of all other nodes.  It really seems like you can't create a
symmetric artemis cluster that works except by using the 231.7.7.7:9876
config.

Has anyone else encountered this issue or ever tried to set up a simple
two-node test cluster using another discovery address/port besides
231.7.7.7 and gotten it to work?  Appreciate any suggestions anyone can
offer.

Thanks,

-a

Re: symmetric cluster UPD other than on 231.7.7.7:9876

[Justin Bertram <jb...@apache.org>]

There's really nothing special about 231.7.7.7:9876. It's just a normal
multicast address & port. Cluster broadcast and discovery should work fine
with any other multicast address & port assuming the environment allows the
communication. Have you verified that the UDP packets are actually making
it from one node to another with something other than 231.7.7.7:9876?

Can you paste the broadcast and discovery group configuration from a
configuration which is not using 231.7.7.7:9876?

Have you tried using 231.7.7.7 but with a different port (e.g. 9877)? If
so, what was the result?


Justin

On Tue, Apr 21, 2020 at 9:46 PM Aaron Dutenhoefer <aa...@gmail.com> wrote:

> I'm using Artemis ActiveMQ v2.11.0 currently - have been using it since 1.x
> days.  I have three environments:  Integration, QA, and Production.  In
> integration, I've got a standalone broker instance.  In QA I'm running a
> two-node symmetric cluster using UDP discovery.  I followed the docs and
> that worked fine.  In Production I'm trying to set up a four-node symmetric
> cluster.
>
> When I set up my first production cluster node, I started it up and it
> immediately joined my QA cluster.  This was expected since the config still
> was using 231.7.7.7:9876 as the discovery address/port.  (My Int/Qa/Prod
> VMs are in the same datacenter/VLAN so UDP traffic routes across them)   I
> then reconfigured the settings using another UDP address - Only to find the
> other nodes configured using the new multicast/UDP address:port never
> connect or discover each other.
>
> I've double-and-triple-checked my configs and even tried using the examples
> and updating my working QA cluster to use another multicast addr (like
> 231.7.7.8 for instance).  Any other address causes the cluster to lose
> sight of all other nodes.  It really seems like you can't create a
> symmetric artemis cluster that works except by using the 231.7.7.7:9876
> config.
>
> Has anyone else encountered this issue or ever tried to set up a simple
> two-node test cluster using another discovery address/port besides
> 231.7.7.7 and gotten it to work?  Appreciate any suggestions anyone can
> offer.
>
> Thanks,
>
> -a
>