Achieving 100% consitency

Jeremiah Peschka jeremiah.peschka at gmail.com
Tue Aug 30 15:57:49 EDT 2011


When I don't need much horsepower, I use 3 local VMs. When I need horsepower, I fire up a few EC2 instances. Either way, once the instances are online, I put the cluster into a variety of states, run some tests, and then shut down my cluster. 

Admittedly, a lot of my time working with Riak is spent writing CorrugatedIron, so I don't spend as much time dealing with the inconsistencies, node gossip, etc. I just try to figure out how developers using .NET are going to want to interact with Riak and then figure out how to turn that into a PBI call. Well, that and make sure that CorrugatedIron is throwing errors to the .NET APIs that don't hide the real problem that Riak is seeing.
---
Jeremiah Peschka - Founder, Brent Ozar PLF, LLC
Microsoft SQL Server MVP

On Aug 30, 2011, at 12:50 PM, Les Mikesell wrote:

> On Tue, Aug 30, 2011 at 2:33 PM, Jeremiah Peschka
> <jeremiah.peschka at gmail.com> wrote:
>> That's the advice I usually give and the exact same reasons I give it.
>> 
>> In an off-list email (as Jonathan knows), I equated it to saying "You should never do RDBMS development without a big enterprise grade server with $500k of SAN behind it." I typically say to build a dev environment that is as close to production as will give you production-like performance. Putting 3 nodes on my crappy stock laptop HDD isn't going to give me anything remotely like a production environment (well… there is always EC2 performance).
>> 
>> We see these questions a lot on the list and I don't feel that telling clients "Well, it's gonna suck until you get it into production" isn't the best possible answer. A lot of people seem to be developing on 3-node clusters made out of Mac Minis, HP micro servers, and EC2; but even more people seem to be developing on laptops and wondering why they're seeing horrible performance.
> 
> How do you test the inconsistent states you will see in production
> when you develop with a one-node cluster?  Or, for that matter on a
> 3-node cluster on the same box where partitioning isn't going to
> happen without some unusual intervention to emulate it?
> 
> -- 
>  Les Mikesell
>   lesmikesell at gmail.com





More information about the riak-users mailing list