reducing n_val

Chris Meiklejohn cmeiklejohn at basho.com
Sun Jul 21 20:43:56 EDT 2013


On Fri, Jul 19, 2013 at 7:35 AM, Simon Effenberg
<seffenberg at team.mobile.de>wrote:

> On Tue, 16 Jul 2013 10:31:07 -0400
> Chris Meiklejohn <cmeiklejohn at basho.com> wrote:
>
> > On Tue, Jul 16, 2013 at 10:02 AM, Simon Effenberg <
> seffenberg at team.mobile.de
> > > wrote:
> >
> > > Changing the n_val setting in the default bucket properties will only
> > > affect new PUT operations moving forward. Older replicas will not be
> > > removed, and the AAE process will not purge older, unneeded replicas.
> > >
> > > Is there a way to get rid of this? Or only by
> > > decommissioning/re-commissioning a node?
> >
> >
> > There isn't a good solution for doing this.  Can I ask why you are
> changing
> > the default n_val for this particular bucket?
>
> because we have some storage problems (it turns out that we have to
> store much more data and at the moment we have not that amount of disk
> space).
>

I would recommend against reducing this value, as you're going to lose
availability
and could risk potential data loss in the event of node failures.


> So is it possible or should it definitely be avoided? Also what is with
> increasing the n_val? Would this lead to many read-repairs through the
> AAE? Or what happens then?
>

If you raise the n_val, read repair and active anti-entropy will handle
creating the
missing replicas.

- Chris
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.basho.com/pipermail/riak-users_lists.basho.com/attachments/20130721/eb62c918/attachment.html>


More information about the riak-users mailing list