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

Re: [libvirt] [PATCH 2/5] virsh: Fix semantics of --config for "update-device" command

On 03/27/13 16:22, Laine Stump wrote:
On 03/21/2013 12:42 PM, Peter Krempa wrote:
The man page states that with --config the next boot is affected. This
can be understood as if _only_ the next bood was affected. This isn't
true if the machine is running.

Are you certain of that? My understanding was that when --config was
specified, *only* the persistent config should be changed, but not the
live state, regardless of whether or not the domain is running.
Otherwise, there is no way to change just the persistent config of a
running domain.

Yes, that's why I'm doing this. Before this patch --config meant "modify the persistent config along with the live change" and there was no way to use this virsh command to change the next-boot config if the domain was running.

Or is your explanation incorrect, and the code correct?

Here is what I *thought* was the meaning of these options:

--config - only change the persistent config, but not the
            live state. The command should fail for a transient

--live   - only change the live state, but not the persistent config.
            The command should fail for a domain that isn't running.

--current - useless (really, I mean that) because its meaning is different
             depending on whether or not the domain is running

Yes those we understand in the same way.

--persistent - deprecated synonym for --config

This used to be the synonym for config, but this might be undestood as the option you are missing later on ...

no option - also useless because it means the same thing as --current

What's missing: a way to say "change both live state and persistent
config as appropriate"

Well, this patch would abuse the --persistent flag for this purpose. It adds _CONFIG always and _LIVE in case the domain is up.

After discussing it on the list, in the name of consistency I actually
very reluctantly implemented this same logic for virsh net-update even
though I thought it was terrible. Did I get it wrong?

No, you've got it correct. Well, except for --persistent but I'm less sure about that than you.

So what is the *real* meaning of each of these options? (and are you
sure you're not changing the meaning of any of them?)

Except for --persistent, the meaning seems to be well defined now and used appropriately in new places. The problem is with the legacy API's that didn't take up on this approach yet.

libvir-list mailing list
libvir-list redhat com

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