[Starlingx-discuss] [Starlingx-disscuss][ovs-dpdk] Reboot instance with physical pci-passthrough NIC causes OVS restart

张鲲鹏 zhang.kunpeng at 99cloud.net
Thu Jul 18 10:09:59 UTC 2019


Hi Xu,Chenjie

I have tried to create VM with 2 pci_passthrough network ports without DPDK, there was the same problem when I rebooted it.
Also, it was same when I reboot the VM with 2 SR-IOV VFs.
Do you have any ideas to debug this problem?

Thanks 
Kunpeng  

> On Jul 17, 2019, at 14:59, Xu, Chenjie <chenjie.xu at intel.com> wrote:
> 
> Hi Kunpeng,
> Maybe you can use SR-IOV and passthrough the VF which has similar performance to physical NIC to the VM. And then you can use DPDK inside the VM with the VF. <>
>  
> Sorry, I don’t have easy way to disable DPDK in stx1.0. The following command is used for stx2.0 which is still in progress:
> system modify --vswitch_type none
>  
> Best Regards,
> Xu, Chenjie
>  <>From: 张鲲鹏 [mailto:zhang.kunpeng at 99cloud.net] 
> Sent: Tuesday, July 16, 2019 5:40 PM
> To: Xu, Chenjie <chenjie.xu at intel.com>
> Cc: starlingx-discuss at lists.starlingx.io
> Subject: Re: [Starlingx-discuss] [Starlingx-disscuss][ovs-dpdk] Reboot instance with physical pci-passthrough NIC causes OVS restart
>  
> Hi Chenjie,
>  
> Well, I will try the network topology as you said. But passthrough NIC with DPDK is our customer’s requirement.
> And do you have some easy ways to disable dpdk of openvswitch in stx1.0? 
> I had tried to execute “system modify --vswitch_type none” before “system host-unlock controller-0", but it doesn’t work well.
>  
> Thanks 
> Kunpeng
>  
> On Jul 16, 2019, at 16:49, Xu, Chenjie <chenjie.xu at intel.com <mailto:chenjie.xu at intel.com>> wrote:
>  
> Hi Kunpeng,
> When you reboot the VM with two physical pci-passthrough NICs, ovs-vswtichd is restarted and the interfaces and bridges are down. The virtual networks used by the VMs are based on these interfaces and bridges. So other VMs will lost connections.
>  
> Typically you should not pass through a PCI device which is bound to DPDK to the VM. Are the 2 network ports which is used to passthrough to the VM bound to DPDK? If so, could you please try OVS-DPDK with the following network topology:
> 2 network port without DPDK      >      VM
> 2 network port with DPDK             >      Data Network
> 1 network port without DPDK       >      OAM
>  
> Best Regards,
> Xu, Chenjie
>  
> From: 张鲲鹏 [mailto:zhang.kunpeng at 99cloud.net <mailto:zhang.kunpeng at 99cloud.net>] 
> Sent: Tuesday, July 16, 2019 3:54 PM
> To: starlingx-discuss at lists.starlingx.io <mailto:starlingx-discuss at lists.starlingx.io>
> Subject: [Starlingx-discuss] [Starlingx-disscuss][ovs-dpdk] Reboot instance with physical pci-passthrough NIC causes OVS restart
>  
> Hi guys,
>  
> Recently I got a strange problem in StarlingX. When I reboot one VM with two physical pci-passthrough NICs, then all of VMs cannot be connected. I lost connections with all VMs and also the VMs lost each others. 
>  
> Below is the StarlingX environment.
>  
> 1. stx1.0 version, bootimage[1] 
> 2. Simplex deployment
> 3. 5 Network ports. Only one don’t support DPDK,and it is used to OAM Network. In the rest, two are used to data network, and another two are used to passthrough to a VM.
> 4. The VM was attached two more virtual networks. I have tested the case of attaching one virtual net, it was no problem. 
>  
> When I reboot the VM, something were happened. The interfaces and bridges were down, all the virtual dhcp services were down and ovs-vswitchd was restarted. But when I up the interfaces and dhcp services and reboot the other VMs, I have got the connections with them again.
>  
> It’s ok when to reboot the VM without physical NIC. We think it may be caused by ovs-dpdk, so we stop to use ovs-dpdk and start the ovs manually, the problem was gone.
>  
> I cannot understand the problem, anybody could give me some comments for it? Thanks a lot.
>  
> [1] http://mirror.starlingx.cengn.ca/mirror/starlingx/r2018.10/centos/2018.10.0/outputs/iso/bootimage.iso <http://mirror.starlingx.cengn.ca/mirror/starlingx/r2018.10/centos/2018.10.0/outputs/iso/bootimage.iso>
>  
> Kunpeng
> _______________________________________________
> Starlingx-discuss mailing list
> Starlingx-discuss at lists.starlingx.io <mailto:Starlingx-discuss at lists.starlingx.io>
> http://lists.starlingx.io/cgi-bin/mailman/listinfo/starlingx-discuss <http://lists.starlingx.io/cgi-bin/mailman/listinfo/starlingx-discuss>
>  
> _______________________________________________
> Starlingx-discuss mailing list
> Starlingx-discuss at lists.starlingx.io
> http://lists.starlingx.io/cgi-bin/mailman/listinfo/starlingx-discuss

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.starlingx.io/pipermail/starlingx-discuss/attachments/20190718/a3f4598d/attachment-0001.html>


More information about the Starlingx-discuss mailing list