fuel-dev team mailing list archive

[Jesse Pretorius <jesse.pretorius@xxxxxxxxx>]

On 24 June 2014 12:50, Dmitriy Novakovskiy <dnovakovskiy@xxxxxxxxxxxx>
wrote:

> I've just realized that CPU overcommitment setting is defined not for
> Controllers but for Compute nodes, and may differ on per-node basis. So,
> the 2nd "harm" example that I exposed a few emails ago (with failed and
> redeployed controller resulting in inconsistent overcommitment behavior) is
> irrelevant.
>

I don't think this is quite right, yet. Right now the ratio is only
implemented in the scheduler - which means it only counts on the
controllers, but there are moves to change this behaviour and move it into
the resource manager. Once that's done it will be on a per-host basis. The
other option at the moment, although I haven't verified this, is that
ratio's can be implemented on a per-aggregate basis.

That said, I do like the idea of a simple selection of options. I would
perhaps also suggest an 'Other' option to allow a user to set whatever
ratio they prefer for the cluster.