[ovirt-users] Falsely detected network device status

[Dan Kenigsberg]

On Thu, Oct 02, 2014 at 12:14:40AM +1000, Lewis Shobbrook wrote:
> Hi All,
>  
> I've encountered a network device that is being falsely detected as down. 
> On adding a new network to a 3.4 hosted engine, one of the two hosts is displaying a component interface as down with the webadmin console.

Which network did you add? On top of which nics? could you share the
setupNetwork command from vdsm.log?

> Ethtool confirms the link status as up and connectivity through the corresponding interfaces of the network has been proven.
> Shortly after adding the network I noted that the node with the falsely detected iface status, was set Non-Operational.
> All running VM's evacuated.
> 
> The engine.log showing...
> INFO  [org.ovirt.engine.core.dal.dbbroker.auditloghandling.AuditLogDirector] (DefaultQuartzScheduler_Worker-33) [2daa7788] Correlation ID: null, Call Stack: null, Custom Event ID: -1, Message: Host hosted_engine_2 moved to Non-Operational state because interfaces 'eth1' are down but are needed by networks 'EVD_DMZ' in the current cluster

Do you have /var/log/vdsm/connectivity.log (it's a recent addition)?
It may should have traces of eth1 going down. Does it? When did it
happen? What else happen at that time on the host?

> 
> 2 other interfaces for ovirtmgmt and another VM network on this same host have long functioned perfectly well.
> 
> When assigning the network to the interface I see the following in messages...
> 
> Oct  1 20:57:39 lx004 kernel: ADDRCONF(NETDEV_UP): eth1: link is not ready
> Oct  1 20:57:39 lx004 kernel: 8021q: adding VLAN 0 to HW filter on device eth1
> Oct  1 20:57:39 lx004 kernel: device eth1 entered promiscuous mode
> Oct  1 20:57:40 lx004 kernel: igb: eth1 NIC Link is Up 100 Mbps Full Duplex, Flow Control: RX
> Oct  1 20:57:40 lx004 kernel: ADDRCONF(NETDEV_CHANGE): eth1: link becomes ready
> Oct  1 20:57:40 lx004 kernel: EVD_DMZ: port 1(eth1) entering forwarding state
> 
> Looks as it should to me, brctl also shows the bridge is as it should be.
> 
> # brctl show
> bridge name	bridge id		STP enabled	interfaces
> ;vdsmdummy;		8000.000000000000	no		
> EVD_DMZ		8000.0025901abc49	no		eth1
> PXE		8000.0007e9a5a197	no		eth3
> ovirtmgmt		8000.0025901abc48	no		eth0
> 
> Not able to see anything relevant in vdsm.log
> 
> vdsClient -s 0 getVdsStats 
> does not list the new network, but instead lists an older deleted one.

Could you elaborate on that? what's reported there exactly? what should
be? Is getVdsCaps up-to-date?

Regards,
Dan.