Thanks - you have confirmed the way that I thought that it works

On 10/ 8/09 02:44 PM, Shawn Walker wrote:

... this is (roughly) what happens if you have
flush-content-on-success set to True:

$ pkg image-update

* pkg creates an update plan

* pkg downloads everything needed for update

* pkg executes plan

* plan execution creates a new boot environment and mounts it

* plan execution performs the update on the new boot environment

* plan execution finishes, purging the contents of /var/pkg/download
_only_ for the new boot environment

This step isn't working, and flush-content-on-success=true is set as
confirmed by pkg property. It's almost as if "true" and "True" are
different. How could we confirm that flush-content-on-success is
observed only if the image-update succeeds in one pass, but is
ignored if (say) beadm fails because it can't find menu.lst and hence
it takes two or more passes? If this is the case, is this a candidate for
a bug report?  Not sure if a) it is legitimate to file bugs against
unsupported features and b)  if it is important enough, given that
beadm will soon be  bug free :-).

* pkg client exits, as the operation is finished

As I said, there are plans to either move /var/pkg/download somewhere
else or make it part of a separate dataset, since, as you pointed out,
destroying a specific dataset is faster than trying to delete 2GB of
individual files.

Can't be too soon for me  - even 32Gb could go awfully quickly and
the delete process  is pretty tedious...
Cheers,
_______________________________________________
indiana-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/indiana-discuss

Reply via email to