On Wed, 16 Aug 2006, Lars Marowsky-Bree wrote:

> On 2006-08-15T09:36:19, Alan Robertson <[EMAIL PROTECTED]> wrote:
>
> > And, unfortunately it would confuse a lot of people, since the IP
> > addresses wouldn't have interface names.  Evidence suggests that would
> > confuse a lot of people -- and other scripts.
>
> But, the addresses being named ethX:Y on Linux has been depreciated
> since 2.4.

Lars: This sounds like an important consideration for Linux systems.
Could you provide us with a background reference (URL) please?  Thanks.

What's the state of play for other OSes?  For Solaris, I'm not aware of
even the existence of anything other than visible "le8:9" format.

> And, the numbers assigned by IPaddr might not even be consistent across
> different nodes.
>
> Scripts which "miss" IPs which don't have a label on Linux are already
> badly broken, the error just occurs less frequently.

So are you saying that Linux good practice is to attach a label on Linux?
A unique label?  Thereby making each logical interface uniquely named?
If so, the result of such good practice would be logically similar to the
traditionally named "eth0:1, eth0:2, eth0:3, ... eth0:<n>", wouldn't it?
(Just with possibly different labels after the ":".)  Further, it would be
wise for our heartbeat scripts to have _default_ (possibly override-able)
behaviour which applied a naming scheme, wouldn't it?


-- 

:  David Lee                                I.T. Service          :
:  Senior Systems Programmer                Computer Centre       :
:                                           Durham University     :
:  http://www.dur.ac.uk/t.d.lee/            South Road            :
:                                           Durham DH1 3LE        :
:  Phone: +44 191 334 2752                  U.K.                  :
_______________________________________________________
Linux-HA-Dev: [email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha-dev
Home Page: http://linux-ha.org/

Reply via email to