[rabbitmq-discuss] clustering on google app engine

[Laing, Michael]

Experiences vary - we cluster across AWS zones in multiple regions and have
not had a partition in production in over a year.

I have experimented with this and the basic rules are: always be sure you
have available compute capacity by over-provisioning your clustered
instances relative to their load; never go into IO wait state (avoid
intense disk usage brought on by swapping or needless message persistence).

Also, avoid the us-east-1 region as it has the most usage, oldest hardware,
highest latencies, and highest incidence of failures.

ml


On Thu, May 29, 2014 at 10:24 PM, Fei Yao <mail2fei at gmail.com> wrote:

> Richard,
> I'm not an expert on GCE. But if there are in the same zone it will not be
> considered WAN. We had some experience in AWS, if it's Multi-AZ it will be
> considered WAN, and we've experience 3 network partitions last year.
>
> Best Regards
>
>
> On Tuesday, May 6, 2014 9:46:39 AM UTC-4, Richard Tier wrote:
>>
>> The docs <https://www.rabbitmq.com/clustering.html> on clustering state
>> that "clustering does not tolerate network partitions well", and states to
>> avoid clustering on WAN.
>>
>> I'm using Google Cloud Engine instances - all of which are in the same
>> zone.
>>
>> Should this setup be considered a WAN?
>>
>
> _______________________________________________
> rabbitmq-discuss mailing list
> rabbitmq-discuss at lists.rabbitmq.com
> https://lists.rabbitmq.com/cgi-bin/mailman/listinfo/rabbitmq-discuss
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.rabbitmq.com/pipermail/rabbitmq-discuss/attachments/20140530/e6f9918c/attachment.html>