[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-users] xentop missing domain state marker

On Fri, 2014-05-02 at 12:41 +0300, Stefan Parvu wrote:
> > My guess is that under load vcpus are being migrated between pcpus and
> > that this "being migrated" state is one which isn't described.
> > 
> makes sense. But we probable need a better state 'm' to show the domains
> are being migrated ?

I think that's a conversation to be had with the relevant maintainers on
xen-devel. It could be argued that "not running due to being migrated"
is just a specific form of "blocked" and can just be reported to the
tools as such.

> > "------" pretty much is "NA".
> > 
> No it is not. If you have raw data collected from xentop / xenrec etc
> and you try to do some numerical analysis with that you are doomed.
> Or we have a valid state, say 'm' or we need to mark it as NA. Tools
> like Mathematica, R etc will digest NA. They wont digest ------ or worse
> this could lead some silent other bugs. Like in my case xenrec.

I don't follow, surely you need to massage any of these states whether
it is ---r---, --s---- or ------- in to something which these tools can
understand. So just massages ------ into NA.

> > You could propose a patch against
> > xen.git/xen/common/domctl.c:getdomaininfo() to consider other states the
> > VPF_blocked as blocked, (list is in xen/include/xen/sched.h) or
> > introduce a new state to be propagated up to the tools (and patch the
> > tools to consume it).
> > 
> > http://wiki.xen.org/wiki/Submitting_Xen_Patches for hints and tips about
> > the patch submission process.
> Thanks a lot. I will do that.

Great, thanks.

>  As soon as Im repeating the tests and read
> over all these things.
> Thanks again for help,

Xen-users mailing list



Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.