Performance increase for Koji builders

[Peter Robinson]

On Mon, Jun 1, 2015 at 1:08 PM, Miroslav Suchý <msuchy at redhat.com> wrote:
> Dne 31.5.2015 v 21:50 Mike McLean napsal(a):
>> Also, some builders do not necessarily have a lot of ram. Some buildroots are large, some sources are extremely large
>> (e.g. texlive). If we do this, it will need to be configurable, probably per-package or per-tag.
>
> The devils in detail is that you do *not* need to have a lot of ram. Just a lot of swap disk.

And a lot of swap disk can cause the IO issues that you mention and
ultimately lead to performance problems.

> E.g. in Copr we had reserved 50 GB for mock chroot. And now instead of that, we now have 50 GB swap.
> Therefore you never ever can have oversized tmpfs, because size of tmpfs is 50 GB, but space for that is 5GB RAM + 50 GB
> swap.

Yes, but it still filters down to IOPs on the disk one way or the other.

> So you consume the same resource. And you get big performance increase.

What package set did you test this against?