Kvm network freeze

Kvm network freeze

[Arthur CLEMENT]

Hi,

I'm experiencing issues with libvirt/kvm on Debian lenny. I have two
servers (32 and amd64) with kvm and openvz and each time I start or
destroy a kvm domain the network freeze during 20 or 30 seconds (even my
openvz vms lost their network).
I think it might come from the bridge config, but this is just a thought.

Any help would be most welcome


Some parameters :

KVM Version: 72+dfsg-5~lenny4
qemu Version: 0.9.1-10lenny1
libvirt : Version: 0.7.6-1~bpo50+1


(vnet come from openvz)
# brctl show
bridge name     bridge id               STP enabled     interfaces
br0             8000.00fff11d455f       no              bond0
                                                        vnet0
                                                        vnet1

/etc/network/interface
# The loopback network interface
auto lo
iface lo inet loopback

# bond0
auto bond0
iface bond0 inet manual
       slaves eth0 eth1
       metric 1
       up ifconfig $IFACE 0.0.0.0 up || true
# br0
auto br0
iface br0 inet static
        address 10.30.15.126
        netmask 255.255.255.0
        gateway 10.30.15.254
        metric 0
        bridge_ports bond0





Dmesg output

[195651.085304] br0: port 2(vnet0) entering disabled state
[195651.141339] device vnet0 left promiscuous mode
[195651.141339] br0: port 2(vnet0) entering disabled state
[197072.483514] kobject_add_internal failed for lo with -EEXIST, don't try
to register things with the same name in the same directory.
[197072.483608] Pid: 28725, comm: libvirtd Not tainted
2.6.26-2-openvz-amd64 #1
[197072.483651]
[197072.483654] Call Trace:
[197072.483741]  [<ffffffff8030f8a9>] kobject_add_internal+0x13f/0x17e
[197072.483791]  [<ffffffff8030fce6>] kobject_add+0x74/0x7c
[197072.483839]  [<ffffffff8029bb16>] ____cache_alloc_node+0x6d/0x106
[197072.483897]  [<ffffffff8029f3ac>] percpu_populate+0x77/0x9e
[197072.483941]  [<ffffffff8030e465>] __next_cpu+0x19/0x26
[197072.483981]  [<ffffffff8030f700>] kobject_get+0x12/0x17
[197072.484024]  [<ffffffff80377a8d>] get_device+0x17/0x1f
[197072.484030]  [<ffffffff803780ae>] device_add+0x9b/0x53f
[197072.484030]  [<ffffffff8030f4f2>] kobject_init_internal+0x12/0x2c
[197072.484030]  [<ffffffff803b1b5a>] register_netdevice+0x282/0x358
[197072.484030]  [<ffffffff803b1c69>] register_netdev+0x39/0x46
[197072.484076]  [<ffffffff8037e7c9>] loopback_net_init+0x3e/0x75
[197072.484076]  [<ffffffff803ae23a>] setup_net+0x85/0xe6
[197072.484076]  [<ffffffff803aea48>] copy_net_ns+0x49/0xb5
[197072.484076]  [<ffffffff8024ad69>] create_new_namespaces+0x125/0x1da
[197072.484076]  [<ffffffff8024b012>] copy_namespaces+0x66/0x9c
[197072.484076]  [<ffffffff802336b8>] copy_process+0xa05/0x11fc
[197072.484092]  [<ffffffff80233fac>] do_fork_pid+0xfd/0x2aa
[197072.484137]  [<ffffffff80221f52>] do_page_fault+0x5c9/0x9ab
[197072.484187]  [<ffffffff802b1db5>] dput+0xc6/0x14f
[197072.484242]  [<ffffffff8020c00a>] system_call_after_swapgs+0x8a/0x8f
[197072.484286]  [<ffffffff8023416a>] do_fork+0x11/0x13
[197072.484325]  [<ffffffff8020c397>] ptregscall_common+0x67/0xb0
[197072.484392]
[199820.431786] device vnet0 entered promiscuous mode
[199820.473424] br0: port 2(vnet0) entering learning state
[199844.086318] vnet0: no IPv6 routers present
[199857.512040] br0: topology change detected, propagating
[199857.512075] br0: port 2(vnet0) entering forwarding state
[201819.368519] device vnet1 entered promiscuous mode
[201819.467978] br0: port 3(vnet1) entering learning state
[201837.971816] vnet1: no IPv6 routers present
[201843.849778] apic write: bad size=1 fee00030
[201843.849903] Ignoring de-assert INIT to vcpu 0
[201844.524922] Ignoring de-assert INIT to vcpu 0
[201846.852021] br0: topology change detected, propagating
[201846.852065] br0: port 3(vnet1) entering forwarding state
[202036.665085] br0: port 3(vnet1) entering disabled state
[202038.731228] device vnet1 left promiscuous mode
[202038.731272] br0: port 3(vnet1) entering disabled state
[202066.345845] device vnet1 entered promiscuous mode
[202066.565852] br0: port 3(vnet1) entering learning state
[202070.052758] apic write: bad size=1 fee00030
[202070.052758] Ignoring de-assert INIT to vcpu 0
[202070.828029] Ignoring de-assert INIT to vcpu 0
[202088.866811] vnet1: no IPv6 routers present
[202101.905984] br0: topology change detected, propagating
[202101.906029] br0: port 3(vnet1) entering forwarding state
[203602.877200] apic write: bad size=1 fee00030
[203602.877200] Ignoring de-assert INIT to vcpu 0
[203603.948087] Ignoring de-assert INIT to vcpu 0
[204019.458601] br0: port 2(vnet0) entering disabled state
[204019.572050] device vnet0 left promiscuous mode
[204019.572100] br0: port 2(vnet0) entering disabled state
[207893.673407] device vnet0 entered promiscuous mode
[207894.215321] br0: port 2(vnet0) entering learning state
[207933.980412] vnet0: no IPv6 routers present
[207946.360943] br0: topology change detected, propagating
[207946.360990] br0: port 2(vnet0) entering forwarding state

Re: Kvm network freeze

[Thomas Løcke]

On Wed, Apr 21, 2010 at 1:32 PM, Arthur CLEMENT <aclement@linagora.com> wrote:
> I'm experiencing issues with libvirt/kvm on Debian lenny. I have two
> servers (32 and amd64) with kvm and openvz and each time I start or
> destroy a kvm domain the network freeze during 20 or 30 seconds (even my
> openvz vms lost their network).
> I think it might come from the bridge config, but this is just a thought.
>
> Any help would be most welcome


I had the same problem on a Slackware host.

The culprit in my case turned out to be the automatic create/destroy
tap functionality  with "-net tap". If I instead did "-net
tap,ifname=SomeTapName,script=no,downscript=no" then it all worked
flawlessly. Of course this solution requires that you know exactly how
many taps you need for a given host, so you can create them upfront,
but in my case that was not a problem.

:o)
Thomas Løcke

Re: Kvm network freeze

[Michael Tokarev]

Arthur CLEMENT wrote:
> Hi,
> 
> I'm experiencing issues with libvirt/kvm on Debian lenny. I have two
> servers (32 and amd64) with kvm and openvz and each time I start or
> destroy a kvm domain the network freeze during 20 or 30 seconds (even my
> openvz vms lost their network).
> I think it might come from the bridge config, but this is just a thought.
> 
> Any help would be most welcome
> 
> 
> Some parameters :
> 
> KVM Version: 72+dfsg-5~lenny4

Upgrade to a version found in backports.org.
Also upgrade your host kernel from the same place.
There's no point at trying to fix an ancient
development snapshot of kvm.

/mjt

Re: Kvm network freeze

[Arthur Clément]

> Arthur CLEMENT wrote:
> > Hi,
> > 
> > I'm experiencing issues with libvirt/kvm on Debian lenny. I have two
> > servers (32 and amd64) with kvm and openvz and each time I start or
> > destroy a kvm domain the network freeze during 20 or 30 seconds (even my
> > openvz vms lost their network).
> > I think it might come from the bridge config, but this is just a thought.
> > 
> > Any help would be most welcome
> > 
> > 
> > Some parameters :
> > 
> > KVM Version: 72+dfsg-5~lenny4
> 
> Upgrade to a version found in backports.org.
> Also upgrade your host kernel from the same place.
> There's no point at trying to fix an ancient
> development snapshot of kvm.
> 
> /mjt

Problem solved with backports (qemu-kvm & kernel), thanks !