Re: non-standard TCP tunings in EC2 images
On 20/07/16 at 11:26 +0200, Thomas Lange wrote:
> >>>>> On Tue, 19 Jul 2016 23:17:15 +0200, Lucas Nussbaum <firstname.lastname@example.org> said:
> > I noticed because #830353, #830452 and #831249 are triggered by this.
> > One could argue that those test suites are a bit fragile, but on the
> > other hand, I would expect an image labelled as "Official Debian" on
> > https://wiki.debian.org/Cloud/AmazonEC2Image to not differ from default
> > Debian configuration in subtle ways like this.
> As a user I would expect that an Official Debian image works like a
> charm on my cloud. As a user, I do not care about FTBS bugs, these are
> bugs a developer cares about, and should be fixed by adjusting the test
> Using tasksel we already have different flavours of Debian
> installations. Why not see the cloud images as another flavour, which
> besides having some additional packages installed also tune some
> parameters for best practice in a cloud environment.
I'm curious about why the sysctl settings I quoted are considered best
practices in cloud environments, and not in other environments
(including standard server environments)?
I'm not against having a discussion about Debian providing different
kernel settings profiles for different environments. This is something
that could probably be useful (those settings can be configured for a
reason, after all).
However, I don't think that this is something that should be decided
silently by the cloud team on its own, but rather discussed with the
kernel team to find a solution that works for everybody.