Re: Network of Brokers: XAException on Failover

[Tim Bain <tb...@alumni.duke.edu>]

A network of brokers can be used for HA as long as the network topology
will remain fully connected in the face of N failures (for whatever value
of N you choose to support, probably 1) and the clients' failover URIs will
always contain a live broker in the face of the same failure(s).

But you could also consider master/slave pairs using LevelDB as the storage
technology, which doesn't have a single point of failure.

Tim
On Sep 29, 2015 3:45 PM, "mhempleman" <ma...@alstom.com> wrote:

> Maybe I'm not using the network of brokers in the correct manner.  Should a
> network of brokers be used for HA, or just scalability and load balancing?
>
> We are not concerned with the loss of a few messages if one broker fails;
> however, we want to make sure there is not a single point of failure in the
> system (one broker node), and we want the client to failover seamlessly in
> the event of a failure (which is not currently happening... see previous
> message).  I looked into master/slave setups, but each seems to have a
> single point of failure.  The failure points in the shared filesystem and
> shared db configurations are obvious.  Any advice would be greatly
> appreciated.  Thanks!
>
>
>
> --
> View this message in context:
> http://activemq.2283324.n4.nabble.com/Network-of-Brokers-XAException-on-Failover-tp4702355p4702445.html
> Sent from the ActiveMQ - User mailing list archive at Nabble.com.
>

Re: Network of Brokers: XAException on Failover

[Tim Bain <tb...@alumni.duke.edu>]

BTW, I'm not convinced that a different network topology will avoid the
exception you first asked about.  Someone who knows XA transactions (which
is not me, sorry) needs to look at that.
On Oct 1, 2015 7:06 AM, "Tim Bain" <tb...@alumni.duke.edu> wrote:

> A network of brokers can be used for HA as long as the network topology
> will remain fully connected in the face of N failures (for whatever value
> of N you choose to support, probably 1) and the clients' failover URIs will
> always contain a live broker in the face of the same failure(s).
>
> But you could also consider master/slave pairs using LevelDB as the
> storage technology, which doesn't have a single point of failure.
>
> Tim
> On Sep 29, 2015 3:45 PM, "mhempleman" <ma...@alstom.com>
> wrote:
>
>> Maybe I'm not using the network of brokers in the correct manner.  Should
>> a
>> network of brokers be used for HA, or just scalability and load balancing?
>>
>> We are not concerned with the loss of a few messages if one broker fails;
>> however, we want to make sure there is not a single point of failure in
>> the
>> system (one broker node), and we want the client to failover seamlessly in
>> the event of a failure (which is not currently happening... see previous
>> message).  I looked into master/slave setups, but each seems to have a
>> single point of failure.  The failure points in the shared filesystem and
>> shared db configurations are obvious.  Any advice would be greatly
>> appreciated.  Thanks!
>>
>>
>>
>> --
>> View this message in context:
>> http://activemq.2283324.n4.nabble.com/Network-of-Brokers-XAException-on-Failover-tp4702355p4702445.html
>> Sent from the ActiveMQ - User mailing list archive at Nabble.com.
>>
>