Change of vdsStats

[Dan Kenigsberg]

On Tue, Jul 09, 2013 at 04:13:27AM -0400, Moti Asayag wrote:
> 
> 
> ----- Original Message -----
> > From: "Dan Kenigsberg" <danken at redhat.com>
> > To: "Petr Sebek" <psebek at redhat.com>, "Moti Asayag" <masayag at redhat.com>
> > Cc: arch at ovirt.org
> > Sent: Monday, July 8, 2013 11:07:49 PM
> > Subject: Re: Change of vdsStats
> > 
> > On Mon, Jul 08, 2013 at 05:25:35AM -0400, Petr Sebek wrote:
> > > The bug on bugzilla was for unknown reasons locked. Now it is unlocked for
> > > everybody.
> > > 
> > > ----- Original Message -----
> > > > From: "Petr Sebek" <psebek at redhat.com>
> > > > To: arch at ovirt.org
> > > > Sent: Thursday, July 4, 2013 1:56:56 PM
> > > > Subject: Change of vdsStats
> > > > 
> > > > Hi,
> > > > 
> > > > I've made patch [1] according to RFE [2]. Basically I just added
> > > > information
> > > > about bridges and vlans in vdsStats. I wanted to start discussion if is
> > > > this
> > > > change needed and suitable? I'm asking because with more statistics comes
> > > > bigger size of output. So I'm asking You if we need this information in
> > > > Engine and about which devices in particular.
> > > > 
> > > > 
> > > > [1] http://gerrit.ovirt.org/#/c/16227/
> > > > [2] https://bugzilla.redhat.com/show_bug.cgi?id=675560
> > 
> > Thanks for your patch, Petr.
> > 
> > It makes vdsm provide all interface information for bridges and vlans -
> > where this information includes mac address, tx/rx rates, 0 in the
> > speed field, and the device state.
> > 
> > Only the latter element is interesting for the purposes of the opened
> > bug.  For example, there's an old patch by Mark Wu suggesting to drop
> > macAddr from the stats of actual nics
> > (http://gerrit.ovirt.org/#/c/13840/); we certainly do not need to report
> > the quite-random mac address of bridge devices.
> > 
> > Similarly, I am not sure that the rxRate of the vlan device has any
> > significance.
> > 
> > Finally, does Engine have plans to collect the state field from bridge
> > devices and report an error if the state is down?
> 
> The engine should move the host to non-operational if the bridge or the vlan 
> are reported from VDSM as down. If it doesn't implemented that way - it is 
> an engine bug.
> 
> > 
> > Can you think of cases, other than manual "ifdown" by an evil admin,
> > where the state if a vlan device is expected to change?

Still, if we cannot think of any use case, besides that of an evil
admin, I think we should CLOSE|WONTFIX the bug.

Dan.