Bug 770153 - vdsm-bootstrap should not set max_cstate=1 to the kernel command line
Summary: vdsm-bootstrap should not set max_cstate=1 to the kernel command line
Keywords:
Status: CLOSED CURRENTRELEASE
Alias: None
Product: oVirt
Classification: Retired
Component: vdsm
Version: unspecified
Hardware: x86_64
OS: Linux
medium
medium
Target Milestone: ---
: 3.2
Assignee: Alon Bar-Lev
QA Contact:
URL:
Whiteboard: virt
Depends On: 770122 770123 bootstrap-rewrite
Blocks:
TreeView+ depends on / blocked
 
Reported: 2011-12-23 16:49 UTC by Yaniv Kaul
Modified: 2016-04-27 02:08 UTC (History)
13 users (show)

Fixed In Version:
Clone Of: 770123
Environment:
Last Closed: 2013-02-15 06:46:26 UTC
oVirt Team: ---
Embargoed:


Attachments (Terms of Use)

Description Yaniv Kaul 2011-12-23 16:49:32 UTC
+++ This bug was initially created as a clone of Bug #770123 +++

+++ This bug was initially created as a clone of Bug #770122 +++

Description of problem:
max_cstate=1 was added to kernel defaults for oVirt Node due to a large portion of hardware having issues w/ time sync in kvm.

As long as hardware has a stable/constant tsc, it should be safe to remove this restriction.  People with older hardware can always manually add it back in.

Need to remove it from vdsm-bootstrap:

[root@rhevm-3 vdsm-bootstrap]# grep -R cstate *
vds_bootstrap.py:            args += ['processor.max_cstate=1']

Comment 1 Dan Kenigsberg 2012-06-25 07:41:16 UTC
This was requested by rhev-3.0 bug 579705, so I'd be keeping max_cstate for rhel6.

http://gerrit.ovirt.org/5658

Comment 2 Yaniv Kaul 2012-06-26 11:42:20 UTC
(In reply to comment #1)
> This was requested by rhev-3.0 bug 579705, so I'd be keeping max_cstate for
> rhel6.

I believe it's no longer needed.
Note that ovirt-node have removed it already - https://bugzilla.redhat.com/show_bug.cgi?id=770123
And the logic to set/remove it - https://bugzilla.redhat.com/show_bug.cgi?id=770123#c2


> 
> http://gerrit.ovirt.org/5658

Comment 3 Dan Kenigsberg 2012-06-26 12:54:25 UTC
I am not at all sure that ovirt-node has satisfied Andy's requirement of https://bugzilla.redhat.com/show_bug.cgi?id=770123#c1

.. we check /proc/cpuinfo flags for constant_tsc and if it's NOT present we add the max_cstate otherwise we leave it off.

This is a logic that vdsmd does not have. Sp this patch is partial.

Comment 4 Alon Bar-Lev 2012-10-23 09:16:10 UTC
I agree this should not be used. Will be removed unless someone has any new input.

Thanks!

Comment 5 Alon Bar-Lev 2012-11-28 08:46:33 UTC
new bootstrap will warn if no constant_tsc.

Comment 6 Itamar Heim 2013-01-16 16:08:36 UTC
3.2 beta built, moving to ON_QA status to allow testing


Note You need to log in before you can comment on or make changes to this bug.