Re: [ovirt-users] [ovirt-devel] Networking fails for VM running on Centos6.7.Works on Centos6.5

[Giorgio Bersano]

2015-11-29 8:59 GMT+01:00 Dan Kenigsberg <danken@xxxxxxxxxx>:
> On Sat, Nov 28, 2015 at 08:10:06PM +0530, mad Engineer wrote:
>> hello all i am having strange network issue with vms that are running on
>> centos 6.7 ovirt nodes.
>>
>> I recently added one more ovirt node which is running centos6.7 and
>> upgraded from centos6.5 to centos6.7 on all other nodes.
>>
>> All VMs running on nodes with centos6.7 as host Operating system fail to
>> reach network gateway,but if i reboot that same host to centos6.5 kernel
>> everything works fine(with out changing any network configuration).
>>
>> Initially i thought it as configuration issue but its there on all nodes.if
>> i reboot to old kernel everything is working.
>>
>> I am aware about ghost vlan0 issue in centos6.6 kernel.Not aware about any
>> issue in centos6.7 Also all my servers are up to date.
>>
>>
>> All physical interfaces are in access mode VLAN connected to nexus 5k
>> switches.
>>
>>
>> working kernel- 2.6.32-431.20.3.el6.x86_64
>>
>> non working kernel- 2.6.32-573.8.1.el6.x86_64

Do you have the possibility to test with
kernel-2.6.32-504.16.2.el6.x86_64 (it's in CentOS Vault now) ?

What I have seen in our environment is that - regarding this problem -
it is the latest correctly working kernel.
There are clear signs of misbehaviours due to changes in the VLAN code
between that kernel and the next one (2.6.32-504.23.4.el6.x86_64). Not
always, it could also depend on the nic driver involved.

Also take a look at https://bugs.centos.org/view.php?id=9467 and
https://bugzilla.redhat.com/show_bug.cgi?id=1263561 .

> Can you provide the topology of your VM network config (vlan, bond, bond
> options, bridge options)? Do you have an IP address on the bridge?
>
> (I have not seen this happen myself)

Dan, did you see this
https://www.mail-archive.com/users@xxxxxxxxx/msg28561.html  thread?

We have seen this bug in oVirt (bond+VLAN+bridges) but also in simple
KVM with VLAN+bridges and with plain servers using VLANS directly over
nic. Therefore I think the topology is almost unrelated and the
problem regards VLAN code.

I would be very interested in
https://bugzilla.redhat.com/show_bug.cgi?id=1264316 but public access
is denied. Do you have access to it? I hope so.

Best regards,
Giorgio.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/