We're not entirely sure of the exact timeline. But here's what we think 
happened. We were building a new cluster, and put a common config on both boxes 
with default n_val = 3. We then manually changed that config on both boxes to 
n_val=1. We restarted riak on one of the nodes, but forgot to restart on the 
other so it's n_val=3 was still in effect. 

That's our best guess.

Dave 

--
Dave Lowell
[email protected]

On Dec 18, 2012, at 3:17 PM, Mark Phillips wrote:

> Hey David,
> 
> So the short answer is that we don't have anything in place that will
> complain wildly if this happens. This, however, should not be the
> case. Before I go file an issue, out of curiosity, how did this
> happen? Did you set the n_val differently in the app.config of each
> node before you started them up?
> 
> Mark
> 
> 
> On Fri, Dec 14, 2012 at 8:03 PM, David Lowell <[email protected]> wrote:
>> I know this is really very boneheaded. But we ended up with n_val being 3 on
>> one of two riak hosts in a dev cluster, and 1 on the other host. I would
>> have expected Riak to howl loudly about this, but it said nothing. In fact,
>> we discovered the problem while debugging app-level issues that resulted
>> from this configuration.
>> 
>> We found that on the two hosts, the same 2i query returned wildly different
>> results. As we investigated we discovered that on the host with n=1, riak
>> was returning each key 3 times. On the host with n=3, each key was in the 2i
>> results once.
>> 
>> Maybe this makes sense at some low level, but I sure would have preferred
>> something in the logs, or even a riak crash to this behavior.
>> 
>> Once we corrected n_val on the errant host, both hosts returned each key 3x.
>> Is it possible for riak to recover from this situation and eventually return
>> the correct 2i results with each key listed only once?
>> 
>> This is a dev cluster, so this isn't critical. But I'd like to understand
>> better whether and how riak can recover from this weird state.
>> 
>> Thanks,
>> 
>> Dave
>> 
>> --
>> Dave Lowell
>> [email protected]
>> 
>> 
>> _______________________________________________
>> riak-users mailing list
>> [email protected]
>> http://lists.basho.com/mailman/listinfo/riak-users_lists.basho.com
>> 

_______________________________________________
riak-users mailing list
[email protected]
http://lists.basho.com/mailman/listinfo/riak-users_lists.basho.com

Reply via email to