On 03/02/2018 02:38 PM, John Ferlan wrote:
On 02/22/2018 05:20 PM, Jim Fehlig wrote:
Long ago in commit dfa1e1dd53 the scheduler weight was accidentally
hardcoded to 1000. Weight is a setting with no unit since it is
relative to the weight of other domains. If no weight is specified,
libxl defaults to 256.
Instead of hardcoding the weight to 1000, honor any <shares> specified
in <cputune>. libvirt's notion of shares is synonomous to libxl's
scheduler weight setting. If shares is unspecified, defer default
weight setting to libxl.
Removing the hardcoded weight required some test fixup. While at it,
add an explicit test for <shares> conversion to scheduler weight.
Signed-off-by: Jim Fehlig <jfehlig@xxxxxxxx>
---
Honoring <shares> specified by the user is certainly desirable,
but I'm not sure about changing the default weight. One problematic
scenario that came to mind: Several domains started by pre-patch
libvirtd with weight=1000, update libvirtd with patch, start more
domains with weight=256. The pre-patch domains unknowingly have
a weight nearly 4 times that of post-patch domains.
Put 'em all on a diet. Don't try to do it overnight though because that
never works and it makes them all hangry... Slowly try to have them
lose weight over the course a few months and in no time they'll be slim,
trim, and less cumbersome. Whether they'll be happy - who knows ;-)
Heh, funny :-).
I'll manage this downstream by ensuring the patch only lands in a new product
release, where no domains will be running when libvirt is installed.
src/libxl/libxl_conf.c | 4 +-
tests/libxlxml2domconfigdata/basic-hvm.json | 2 +-
tests/libxlxml2domconfigdata/basic-pv.json | 2 +-
tests/libxlxml2domconfigdata/cpu-shares-hvm.json | 89 ++++++++++++++++++++++
tests/libxlxml2domconfigdata/cpu-shares-hvm.xml | 39 ++++++++++
tests/libxlxml2domconfigdata/moredevs-hvm.json | 2 +-
tests/libxlxml2domconfigdata/multiple-ip.json | 2 +-
.../libxlxml2domconfigdata/variable-clock-hvm.json | 2 +-
tests/libxlxml2domconfigdata/vnuma-hvm.json | 2 +-
tests/libxlxml2domconfigtest.c | 1 +
10 files changed, 138 insertions(+), 7 deletions(-)
Perhaps this is a "news" worthy type change along with the caveat about
those more weighty domains that were allowed to feed at the trough
without being limited.
I've sent a patch to add a news entry, including a short description of the
potential problem + mitigation
https://www.redhat.com/archives/libvir-list/2018-March/msg01336.html
Regards,
Jim
--
libvir-list mailing list
libvir-list@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/libvir-list
