Re: Second regression due to libxl: Remove linux udev rules (2ba368d13893402b2f1fb3c283ddcc714659dd9b)

[Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>]

On Tue, Aug 18, 2015 at 09:49:43AM +0200, Roger Pau Monné wrote:
> El 12/08/15 a les 16.09, Konrad Rzeszutek Wilk ha escrit:
> > On Fri, Aug 07, 2015 at 04:58:57PM +0200, Roger Pau Monné wrote:
> >> El 07/08/15 a les 16.54, Konrad Rzeszutek Wilk ha escrit:
> >>> Ok. I hadn't run your patch yet. Do you want me to run the latest staging
> >>> instead once more with my test-case?
> >>
> >> Yes please, 40s in my test case seemed to be fine.
> > 
> > I get inconsistent values. Sometimes it works, sometimes it does not.
> > 
> > I will try 120 on my test-bed :-(
> 
> Thanks, this is not unexpected. IMHO the best way to solve this would be
> to make the hotplug timeout settable by the user by adding a new option
> to xl.conf:
> 
> hotplug_timeout=40
> 
> Allowing the user to set it to -1 in order to remove the timeout if desired.
> 
> This however seems too intrusive IMHO at this point in the release, so I
> would argue for just increasing the default timeout to a suitable value.
> Konrad do you have any suggestions?

I've been running with staging for months now and I don't see this problem.

Going forward I would think that making this work as you explained
with the 'hotplug_timeout=-1' would be nice. But then I seem to be the
only one doing it - and I don't see the normal user re-using the same
file or phy to launch tons of guests in parallel.

In my mind this is very much a feature for a developer and as such I
would say less important as a developer I can always just modify the
source code with a higher timeout.

> 
> Roger.
>