Den 8 jan. 2018 08:32 skrev Yedidyah Bar David <didi@redhat.com>:
On Mon, Jan 8, 2018 at 1:36 AM, Sam McLeod <mailinglists@smcleod.net> wrote:
Thank you for the information Dan, Dominik and Didi,

To avoid logging yet another bug for this issue, I've updated bug https://bugzilla.redhat.com/show_bug.cgi?id=1459229 as you've mentioned with the brief of our conversation here.

By the way, it is very useful to name a bonded interface things other than bondXYZ, for example, you might have 6 bonds, each of a different network or native VLAN.
It helps with debugging, troubleshooting and logging if the interface is named after the (native) network, e.g. your iSCSI storage network might have a bond called 'storage', while your management or hypervisor network might have a bond named 'mgmt' then perhaps you have 'data' bond that might have several vlans off it such as 'db' (database), 'dmz', 'staff' etc... depending on how and where you chop your network up.

When I was a sysadmin I used to call my bonds bondFUNCTION.

This way I both had a prefix 'bond' that readily showed it's
a bond, and a suffix showing its function.

IMO oVirt should allow any bond names. If we do decide to limit
them at all, I'd limit only in a negative way - what's not
allowed. E.g. it makes sense to me if we reject prefixes that
are common for non-bonds (eth, en, wl, br etc), but even that
I am not sure is so important.

I agree, while at the same time, I don't understand why you would want to "label" a bond. Even if it's untagged, since you can label the bridge that's ontop of it? So even if the bond is called "bond0", the bridge can be called "data", or whatever, to ease diagnozing etc.

What if it's tagged? Would you want to label both bond, vlan and bridge? Like "foobond", "barvlan" and "bazbridge"?

/K

 

On 7 Jan 2018, at 6:08 pm, Yedidyah Bar David <didi@redhat.com> wrote:

On Fri, Jan 5, 2018 at 7:44 AM, Dan Kenigsberg <danken@redhat.com> wrote:
On Thu, Jan 4, 2018 at 5:50 AM, Sam McLeod <mailinglists@smcleod.net> wrote:
> I'm having a problem where when setting up hosted engine deployment it fails
> stating that the selected bond name is bad.
>
> "code=25, message=bad bond name(s): mgmt)"
>
> - Is there a problem similar to
> https://bugzilla.redhat.com/show_bug.cgi?id=1519807 that's known?

Please note that this is just but one bug in a series/tree of
related bugs, some of which are open. If you decide to follow
Dan's suggestion, perhaps reuse one of the others, or perhaps
even better - open a new one, and eventually one or more will
be closed as duplicate of one or more of the others. Sadly,
not all of them link properly to each other, and at least one
which was fixed caused another bug, so the fix was reverted.
See also e.g. all of the discussion in:

https://bugzilla.redhat.com/show_bug.cgi?id=1459229

 
> - If it seems to be this bug, is it preferred that I simply update the
> existing, closed issue as I have done, or open a new bug?
>
> --
> Sam McLeod
> https://smcleod.net
> https://twitter.com/s_mcleod

I see that you are trying to use a bond interface named "mgmt".
To avoid confusion while debugging a system, Vdsm has opted to allow
only bond names starting with "bond" followed by one or more decimal
digits. Anything else is considered "bad bond".

I prefer keeping the "bond" prefix compulsory, but I'd like to hear
why using different names is useful.

You can reopen this bug, but please move it to vdsm and rename it: it
should be something like "Allow any bondXYZ name for bonds" or "Allow
any bond name" and explain there why it is a good idea.

Dominik, is there an Engine-side limitation on bond names?
_______________________________________________
Users mailing list
Users@ovirt.org
http://lists.ovirt.org/mailman/listinfo/users



--
Didi




--
Didi
_______________________________________________
Users mailing list
Users@ovirt.org
http://lists.ovirt.org/mailman/listinfo/users