activemq-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Tim Bain <tb...@alumni.duke.edu>
Subject Re: [Artemis] Moving all messages from one broker cluster to another
Date Thu, 28 Feb 2019 14:55:51 GMT
Justin's question about AWS replacing nodes and not needing to replicate
messages would be applicable if the update strategy being used was rolling
update, with the new hosts being launched into the same cluster. But with
blue-green, you're making a second cluster, so you have to ensure that
messages are transferred somehow.

In ActiveMQ 5.x, you'd probably want to network the two clusters together
into a network of brokers, then switch all producers and consumers to the
new cluster, and allow all messages to drain off the old broker. That
relies on having consumers on all queues that have messages, and having no
offline durable consumers on topics, and no scheduled messages; if any of
those were applicable, things would get more complicated. I don't know if
Artemis has a better solution than that for this migration problem, but
it's something 5.x never really created a good solution for and it would be
a good thing for Artemis to create a solution for if one doesn't already
exist.

Tim

On Wed, Feb 27, 2019, 7:04 AM Justin Bertram <jbertram@apache.org> wrote:

> Take a look at the export/import functionality available from the CLI.
>
> Out of curiosity, will AWS not restart a node if it goes down and if so
> wouldn't that mitigate the need for a backup?
>
>
> Justin
>
> On Wed, Feb 27, 2019 at 4:12 AM <jo.stenberg@offilive.com> wrote:
>
> > Hi,
> >
> > we are using Artemis 2.6 clusters in AWS where a cluster consists of a
> > master and one backup server. For upgrades we are using a blue/green
> > deployment scheme where we setup a new cluster on two new server
> instances,
> > perform functional tests etc and once done switch the loadbalancer in
> front
> > of Artemis from the old cluster to the new cluster.
> >
> > Now we are trying to find an easy way to migrate all messages from the
> old
> > cluster to the new cluster.
> >
> > I looked at Core Bridges but they require manual configuration for each
> > source/target address. We however have a lot autocreated
> addresses/queues.
> >
> > Any ideas?
> >
> > Thanks,
> > Jo
> >
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message