UCA & HWE kernel
James Page
james.page at ubuntu.com
Wed May 15 13:03:21 UTC 2013
Hi David
On 15/05/13 03:25, David Medberry (Med) wrote:
> Does the migration to the HWE kernel in Precise (at some point) effect
> the UCA work? I understand that after the next point release, the HWE
> kernel will likely be the default.
>
> Ref:
> http://goo.gl/erP8b (UDS Session on 12.04.x LTS with HWE Kernel)
>
> This session was just ahead of the UCA session (and I was unable to
> attend the the first session so missed it but replayed it later in the
> day.)
>
> There was a bit more discussion in the Kernel Misc (but this also took
> place without much representation from the cloud/server folks though
> Brad Figg did at least carry the discussion into the meeting.) Leann is
> looking for requirements from cloud/server.
Yeah - I wanted to make that meeting but I was participating on one in
the cloud track. We have a round-table tomorrow - I'll ensure its
raised there.
> My concern is
> a) What will we be testing in UCA (always the original precise LTS
> kernel stream)
That's what we test with right now but there are some features in later
kernels not related to hardware enablement that will appeal to cloud
archive users. For example, the Ceph RBD driver in the 3.5 kernel is
much more solid that in 3.2; likewise later kernels have support for
VXLAN (for which support should land in OpenStack during Havana).
> b) What will we be suggesting "new" users of UCA use
I think that depends on "I understand that after the next point release,
the HWE kernel will likely be the default." i.e. if all new installs and
upgrades will switch to the HWE kernel, then we make that
recommendation; however if that is not the case we should stick with 3.2
kernel as shipped in 12.04 as the base recommendation.
> c) Are we going to be locking out newer hardware (ie, dell/HP servers)
> if we only test/qual UCA with the original.
Hopefully we can avoid that.
> This sounds like another potential test matrix explosion (and maybe it
> is being covered by already testing in Raring/Saucy + Precise) but I
> thought I'd air the question.
Its covered at-least partially by the fact that we test both for the
main distro and for the UCA for precise - but I think we need to beef up
coverage. Right now this is a little awkward to support, but with the
switch to OpenStack based testing of OpenStack, we should be able to
support this more easily.
Thanks for raising this.
Cheers
James
--
James Page
Ubuntu Core Developer
Debian Maintainer
james.page at ubuntu.com
More information about the ubuntu-server
mailing list