[ejabberd] Load balancing and clustering

Raoul Duke rduke496 at gmail.com
Mon Dec 25 03:08:12 MSK 2017


Thanks again for your feedback.  One final question on this broad topic: if
I had access to a large piece of hardware.  Lets sat 64 core with 256GB RAM.

would there be any value add in runing a lot of virtual appliance on that
host each running an ejabberd cluster pool member.

*or*

would it be still better to have a single host using all the resources and
a single ejabberd instance?

I'm aware that ejabberd and erlang can use as many cores as you can throw
at it.

assuming the single ejabberd was stable I'm struggling to think of a
situation where multiple pool members would be a win.  or am I wrong and
does segmenting the hardware into virtual appliances each with a clustered
ejabberd pool member make sense?

I'd appreciate your thoughts on this.

Thanks.
RD,




On Wed, Dec 20, 2017 at 12:38 PM, Evgeny Khramtsov <xramtsov at gmail.com>
wrote:

> Wed, 20 Dec 2017 11:55:36 +0000
> Raoul Duke <rduke496 at gmail.com> wrote:
>
> > I'm assuming that terminating SSL is problematic at the load balancer
> > because of STARTTLS?  So would it be tcp load balancing with TLS still
> > handled by the pool members?
>
> It's possible to do Direct-TLS if your clients support XEP-0368:
> https://xmpp.org/extensions/xep-0368.html
>
> > Does anyone have a link (or can paste) a sample haproxy config that
> > you use in front of a cluster?  I just don't want to learn lessons
> > the hard way.
>
> https://blog.onefellow.com/post/76702632637/haproxy-and-ejabberd
> Not sure how fresh this article is, though.
> _______________________________________________
> ejabberd mailing list
> ejabberd at jabber.ru
> http://lists.jabber.ru/mailman/listinfo/ejabberd
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.jabber.ru/pipermail/ejabberd/attachments/20171225/800b092b/attachment.html>


More information about the ejabberd mailing list