Re: [Xen-devel] [OSSTEST Nested PATCH v8 6/7] Compose the main recipe of nested test job

[Robert Hu <robert.hu@xxxxxxxxxxxxxxxx>]

On Thu, 2015-04-23 at 12:05 +0100, Ian Jackson wrote:
> Pang, LongtaoX writes ("RE: [Xen-devel] [OSSTEST Nested PATCH v8 6/7] Compose 
> the main recipe of nested test job"):
> > Ian Campbell [mailto:ian.campbell@xxxxxxxxxx]:
> > > The arguments passed to ts-* in sg-run-job should _always_ be just a
> > > bare indent. The ident=foo thing is a special case for standalone mode.
> > > 
> > Do you means that the parameter passed to ts-* in sg-run-job should 
> > _always_ be just a bare ident?
> 
> Yes, that's what Ian means.
> 
> > If yes. I think it needs to pass l1's '$gn' which is not a bare ident to 
> > 'ts-debian-hvm-install' for installing L1 guest VM, since the default '$gn 
> > ||= 'debianhvm' if no other '$gn' was passed to this script, but for nested 
> > job we expect a different '$gn' from 'debianhvm', such as 'nestedl1'. So it 
> > need to pass 'host' and 'nestedl1' to  'ts-debian-hvm-install' script. 
> > +    run-ts . = ts-debian-hvm-install + host + nestedl1
> 
> Indeed.
> 
> > For 'ts-nested-setup' script, it need to call 'ts_get_host_guest'
> > which need two parameters 'host' and 'nestedl1' to get $l0 and $l1
> > scope, so it also need to pass 'nestedl1' to this script.
> 
> Yes.
> 
> > For ' ts-guest-destroy', it also need two parameters of 'host' and 
> > 'nestedl1' which is L1's '$gn'.
> 
> If you're intending to destroy the l1, then yes.  If you want to
> destroy theh l2 then you need to pass `nestedl1' and whatever the L2
> guest name/ident is.
Here comes a common issue we need to settle down:
To not break the principle that we shall only pass in ident in recipe,
we shall either
1. for l1 and l2, identical their ident/name/hostname, make this a
principle as well. This shall avoid much complex design but adding one
additional rule to bear in mind.
2. setup $ident --> $name mapping somewhere at the beginning (either
alloc host or make flight, you name it, but somewhere).

I prefer Ian C's suggestion, option 1, that why should we differ the
ident and name for l1 and l2.
> 
> Ian.



_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel