On 10/31/23 03:59, Liqi An wrote:
> Hi ,
>       Ok , I will continue to consult SUSE support on this issue .
> 
> About the new workaround, I hope bond1 work as before ,
> 
> A> Old solution(bond1 in bridge: br-oam will lost when restarting network of 
> host):
> cluster12-b:/etc/sysconfig/network # cat ifcfg-bond1
> DEVICE='bond1'
> BONDING_MASTER='yes'
> BONDING_MODULE_OPTS='mode=active-backup miimon=100 use_carrier=0'
> BONDING_SLAVE0='eth1'
> BONDING_SLAVE1='eth5'
> BOOTPROTO='static'
> BORADCAST=''
> ETHTOOL_OPTIONS=''
> IPADDR=''
> MTU=''
> NAME=''
> NETMASK=''
> NETWORK=''
> REMOTE_IPADDR=''
> STARTMODE='auto'
> USERCONTROL='no'
> BONDING_SKIP_REMOVE_WORKAROUND='yes'
> ZONE=public
> 
> # /usr/bin/ovs-vsctl add-port br-oam bond1 trunk=3932,3933
> 
> Bridge br-oam
>         Port "2.11-SC-2-eth1"
>             tag: 3932
>             Interface "2.11-SC-2-eth1"
>         Port bond1
>             trunks: [3932, 3933]
>             Interface bond1
>         Port "2.11-SC-2-eth2"
>             tag: 3933
>             Interface "2.11-SC-2-eth2"
>         Port br-oam
>             Interface br-oam
>                 type: internal
> 
> 
> 
> 
> B> New solution:
> # ovs-vsctl add-bond br-oam bond1 eth1 eth5 trunk=3932,3933
> 
>    Bridge br-oam
>         Port br-oam
>             Interface br-oam
>                 type: internal
>         Port bond1
>             trunks: [3932, 3933]
>             Interface eth5
>             Interface eth1
>         Port "2.11-SC-2-eth2"
>             tag: 3933
>             Interface "2.11-SC-2-eth2"
>         Port "2.11-SC-2-eth1"
>             tag: 3932
>             Interface "2.11-SC-2-eth1"
> 
>       Would bond1 in solution B> work in the same way as bond1 in solution A> 
> , Especially the work of eth1&eth5 
> as ' BONDING_MODULE_OPTS='mode=active-backup miimon=100 use_carrier=0''? Or 
> more parameter configuration is required in 
> command"# ovs-vsctl add-bond br-oam bond1 eth1 eth5 trunk=3932,3933" ?

By default the bonding mode is active-backup, but the default failure
detection mode is carrier, so you'll need to change that in order to
match the previous configuration.  For example:

  ovs-vsctl set port bond1 other_config:bond-detect-mode=miimon \
                           other_config:bond-miimon-interval=100


> 
> 
> //An
> 
> -----Original Message-----
> From: Ilya Maximets <[email protected]> 
> Sent: Monday, October 30, 2023 7:09 PM
> To: Liqi An <[email protected]>; [email protected]
> Cc: [email protected]; Cheng Chi <[email protected]>; Jonas Yi 
> <[email protected]>; Yawei Lu <[email protected]>
> Subject: Re: [ovs-discuss] Restarting the network triggers the deletion of 
> one ovs port
> 
> On 10/30/23 10:28, Liqi An wrote:
>> Hi ,
>>      Is there any update about this issue , It's been bothering me for 
>> over two weeks,thx~
> 
> Hi.  As you saw in the log, something is calling ovs-vsctl to remove the port 
> from OVS:
> 
> 2023-10-16T13:07:17.668420+08:00 cluster12-b ovs-vsctl: 
> ovs|00001|vsctl|INFO|Called as /usr/bin/ovs-vsctl del-port br-oam bond1
> 
> OVS is not doing that on its own.  You need to find what is calling this 
> command in order to fix the problem.  Likely candidates are network-scripts, 
> NetworkManager or something similar.
> 
> Best regards, Ilya Maximets.
> 
>>
>> //An
>>
>> -----Original Message-----
>> From: Liqi An
>> Sent: Wednesday, October 18, 2023 5:13 PM
>> To: Ilya Maximets <[email protected]>; [email protected]
>> Cc: Cheng Chi <[email protected]>; Jonas Yi 
>> <[email protected]>; Yawei Lu <[email protected]>; IPW AQUA 
>> team (CBC) <[email protected]>
>> Subject: RE: [ovs-discuss] Restarting the network triggers the 
>> deletion of one ovs port
>>
>> Hi ,
>>      We added bond1 to br-oam with below command before:     
>>
>> # ovs-vsctl add-port br-oam bond1 trunk=3932,3933
>>
>>      In my opinion , , once it  add bond1 to the virtual switch 
>> (openvswtich) successfully , this configuration should be saved into ovs db 
>> , What I'm wondering is why the network management on the host actively  
>> triggers the deletion, That's not supposed to be the job of SUSE15-SP5 , The 
>> ovs-vsctl command is used to manually add or delete these configurations 
>> before.
>>
>>
>>      By the way , I tried another way , delete bond1 configuration of 
>> previous host , and add eth1&eth5 as bond1 to openvswitch directly:
>> WA:
>> cluster12-b:/etc/sysconfig/network # ip link set bond1 down 
>> cluster12-b:/etc/sysconfig/network # rm -f 
>> /etc/sysconfig/network/ifcfg-bond1
>> cluster12-b:/etc/sysconfig/network # service network restart 
>> cluster12-b:/etc/sysconfig/network # ip link set eth1 up 
>> cluster12-b:/etc/sysconfig/network # ip link set eth5 up
>>  
>> cluster12-b:/etc/sysconfig/network # ovs-vsctl add-bond br-oam bond1 eth1 
>> eth5 trunk=3932,3933 cluster12-b:/etc/sysconfig/network # ovs-vsctl show 
>> 2e9bf291-50ac-4c3a-ac55-2d590df1880d
>>      Bridge br-oam
>>          Port br-oam
>>              Interface br-oam
>>                  type: internal
>>          Port bond1
>>              trunks: [3932, 3933]
>>              Interface eth1
>>              Interface eth5
>>      ovs_version: "2.14.2"
>> cluster12-b:/etc/sysconfig/network # service network restart 
>> cluster12-b:/etc/sysconfig/network # ovs-vsctl show 
>> 2e9bf291-50ac-4c3a-ac55-2d590df1880d
>>      Bridge br-oam
>>          Port br-oam
>>              Interface br-oam
>>                  type: internal
>>          Port bond1
>>              trunks: [3932, 3933]
>>              Interface eth1
>>              Interface eth5
>>      ovs_version: "2.14.2"
>>
>> After initial testing, this scheme works temporarily and does not lose the 
>> relevant network configuration after restarting the network and host ,But I 
>> not sure  how eth1&eth5 works in this configuration , I hope them work as 
>> before:
>>
>> cluster12-b:~ # cat /etc/sysconfig/network/ifcfg-bond1
>> DEVICE='bond1'
>> BORADCAST=''
>> NETWORK=''
>> USERCONTROL='no'
>> BONDING_SKIP_REMOVE_WORKAROUND='yes'
>> BOOTPROTO='static'
>> STARTMODE='auto'
>> ZONE='public'
>> BONDING_MASTER='yes'
>> BONDING_SLAVE0='eth5'
>> BONDING_SLAVE1='eth1'
>> BONDING_MODULE_OPTS='mode=active-backup miimon=100 use_carrier=0'
>>
>> do you have any idea about this solution or give some advice, thx ~
>>
>> //An
>>
>> -----Original Message-----
>> From: Ilya Maximets <[email protected]>
>> Sent: Tuesday, October 17, 2023 7:34 PM
>> To: Liqi An <[email protected]>; [email protected]
>> Cc: [email protected]; Cheng Chi <[email protected]>; Jonas Yi 
>> <[email protected]>; Yawei Lu <[email protected]>
>> Subject: Re: [ovs-discuss] Restarting the network triggers the 
>> deletion of one ovs port
>>
>> On 10/17/23 07:53, Liqi An wrote:
>>> Hi experts ,
>>>     I simplified the issue repetition steps ,
>>>
>>> cluster12-b: # cat ovs-network.xml
>>> <?xml version="1.0"?>
>>> <network>
>>>     <name>2.11-ovs-network</name>
>>>     <forward mode="bridge"/>
>>>     <bridge name="br-oam"/>
>>>     <virtualport type="openvswitch"/>
>>>     <portgroup name="2.11-vlan-01" default="yes">
>>>     </portgroup>
>>>     <portgroup name="2.11-SC-eth1">
>>>         <vlan>
>>>             <tag id="3932"/>
>>>         </vlan>
>>>     </portgroup>
>>>     <portgroup name="2.11-SC-eth2">
>>>         <vlan>
>>>             <tag id="3933"/>
>>>         </vlan>
>>>     </portgroup>
>>>     <portgroup name="2.11-vlan-all">
>>>         <vlan trunk="yes">
>>>             <tag id="3932"/>
>>>             <tag id="3933"/>
>>>         </vlan>
>>>     </portgroup>
>>> </network>
>>> cluster12-b: # virsh list --all
>>>  Id   Name   State
>>> --------------------
>>>
>>> cluster12-b: # virsh net-list --all
>>>  Name   State   Autostart   Persistent
>>> ----------------------------------------
>>>
>>> cluster12-b: # virsh net-define ovs-network.xml Network 
>>> 2.11-ovs-network defined from ovs-network.xml
>>>
>>> cluster12-b: # virsh net-list --all
>>>  Name               State      Autostart   Persistent
>>> -------------------------------------------------------
>>>  2.11-ovs-network   inactive   no          yes
>>>
>>> cluster12-b: # virsh net-start 2.11-ovs-network Network 
>>> 2.11-ovs-network started
>>>
>>> cluster12-b: # virsh net-list --all
>>>  Name               State    Autostart   Persistent
>>> -----------------------------------------------------
>>>  2.11-ovs-network   active   no          yes
>>>
>>> cluster12-b: # ovs-vsctl show
>>> 2e9bf291-50ac-4c3a-ac55-2d590df1880d
>>>     ovs_version: "2.14.2"
>>> cluster12-b: # ovs-vsctl add-br br-oam
>>> cluster12-b: # ovs-vsctl show
>>> 2e9bf291-50ac-4c3a-ac55-2d590df1880d
>>>     Bridge br-oam
>>>         Port br-oam
>>>             Interface br-oam
>>>                 type: internal
>>>     ovs_version: "2.14.2"
>>> cluster12-b: # ovs-vsctl add-port br-oam bond1 trunk=3932,3933
>>> cluster12-b: # ovs-vsctl show
>>> 2e9bf291-50ac-4c3a-ac55-2d590df1880d
>>>     Bridge br-oam
>>>         Port br-oam
>>>             Interface br-oam
>>>                 type: internal
>>>         Port bond1
>>>             trunks: [3932, 3933]
>>>             Interface bond1
>>>     ovs_version: "2.14.2"
>>> cluster12-b: # date
>>> Tue Oct 17 13:47:02 CST 2023
>>> cluster12-b: # service network restart
>>> cluster12-b: # ovs-vsctl show
>>> 2e9bf291-50ac-4c3a-ac55-2d590df1880d
>>>     Bridge br-oam
>>>         Port br-oam
>>>             Interface br-oam
>>>                 type: internal
>>>     ovs_version: "2.14.2"
>>> cluster12-b: #
>>>
>>> it seems like a common issue . 
>>
>> You have in your log "kernel:  [340557.486796][ T2447] bonding: bond1 is 
>> being deleted", it means something is destroying a bonding device in the 
>> kernel.  And it makes some sense because you're restarting the network.  
>> Does this bond1 interface exist after the network restart?  What creates it 
>> back? network-scripts?  If so, you should check these network-scripts.  And 
>> the port-add command should be somewhere in the same network scripts.
>> And the bridge should likely be created there.  Otherwise, the configuration 
>> you created is not going to be persistent.
>>
>>>
>>> //An
>>>
>>> -----Original Message-----
>>> From: Ilya Maximets <[email protected]>
>>> Sent: Monday, October 16, 2023 6:54 PM
>>> To: Liqi An <[email protected]>; [email protected]
>>> Cc: Cheng Chi <[email protected]>; Jonas Yi 
>>> <[email protected]>; Yawei Lu <[email protected]>; 
>>> [email protected]
>>> Subject: Re: [ovs-discuss] Restarting the network triggers the 
>>> deletion of one ovs port
>>>
>>> On 10/16/23 07:38, Liqi An via discuss wrote:
>>>> Hi experts,
>>>>
>>>>    I am having a very strange problem with matching virtual machines 
>>>> installations with openvswitch. My operating system is suse15-sp4;
>>>>
>>>> cluster12-b:~ # cat /etc/os-release
>>>> /NAME="SLES"/
>>>> /VERSION="15-SP4"/
>>>> /VERSION_ID="15.4"/
>>>> /PRETTY_NAME="SUSE Linux Enterprise Server 15 SP4"/ /ID="sles"/ 
>>>> /ID_LIKE="suse"/ /ANSI_COLOR="0;32"/ 
>>>> /CPE_NAME="cpe:/o:suse:sles:15:sp4"/
>>>> /DOCUMENTATION_URL=https://documentation.suse.com/
>>>> <https://documentation.suse.com/>/
>>>>
>>>> cluster12-b:~ # rpm -qa |grep openvswitch 
>>>> /openvswitch-2.14.2-150400.22.23.x86_64/
>>>>
>>>> cluster12-b:~ # virsh net-list --all
>>>>
>>>> /Name               State    Autostart   Persistent/ 
>>>> /-----------------------------------------------------/
>>>> /2.11-ovs-network   active   yes         yes/
>>>>
>>>> bond1 was used by the VMs:
>>>> …
>>>>    Bridge br-oam
>>>>         Port bond1
>>>>             trunks: [3932, 3933]
>>>>             Interface bond1
>>>>         Port "2.11-SC-2-eth1"
>>>>             tag: 3932
>>>>             Interface "2.11-SC-2-eth1"
>>>>         Port br-oam
>>>>             Interface br-oam
>>>>                 type: internal
>>>>         Port "2.11-SC-2-eth2"
>>>>             tag: 3933
>>>>             Interface "2.11-SC-2-eth2"
>>>>
>>>>      But when I restarted the network service by command: # service 
>>>> network restart , this port bond1 lost in the bridge br-oam , and 
>>>> there are some abnormal log in systemlog, Detailed operation logs 
>>>> are attached
>>>>
>>>> …
>>>> /25302 2023-10-16T13:07:12.708071+08:00 cluster12-b kernel: 
>>>> [340552.475586][ T2447] device eth1 left promiscuous mode/
>>>> /25303 2023-10-16T13:07:12.824022+08:00 cluster12-b kernel: 
>>>> [340552.593298][ T2447] bonding: bond0 is being deleted.../
>>>> /25304 2023-10-16T13:07:12.824045+08:00 cluster12-b kernel: 
>>>> [340552.593393][ T2447] bond0 (unregistering): Released all slaves/
>>>> /25305 2023-10-16T13:07:12.881576+08:00 cluster12-b systemd[1]: 
>>>> Starting Generate issue file for login session.../
>>>> /25306 2023-10-16T13:07:12.905589+08:00 cluster12-b systemd[1]: 
>>>> issue-generator.service: Deactivated successfully./
>>>> /25307 2023-10-16T13:07:12.905662+08:00 cluster12-b systemd[1]: 
>>>> Finished Generate issue file for login session./
>>>> /25308 2023-10-16T13:07:17.668420+08:00 cluster12-b ovs-vsctl: 
>>>> ovs|00001|vsctl|INFO|Called as /usr/bin/ovs-vsctl del-port br-oam
>>>> bond1/
>>>> /25309 2023-10-16T13:07:17.676015+08:00 cluster12-b kernel: 
>>>> [340557.444150][ T2261] device bond1 left promiscuous mode/
>>>> /25310 2023-10-16T13:07:17.720080+08:00 cluster12-b kernel: 
>>>> [340557.486796][ T2447] bonding: bond1 is being deleted.../
>>>> /25311 2023-10-16T13:07:17.720097+08:00 cluster12-b kernel: 
>>>> [340557.486891][ T2447] bond1 (unregistering): Released all slaves/
>>>
>>> IIUC, the 'bond1' is some sort of a kernel bonding device configured 
>>> outside of OVS.  And it is getting removed.
>>> When you restart the network, the system will execute whatever network 
>>> configuration is in your system settings, e.g. stuff from 
>>> /etc/sysconfig/network-scripts, maybe NetworkManager is going to re-apply 
>>> its configuration or netplan, I don't really know what SUSE is using.
>>> So, you should look in these places for things that manage the bond1 
>>> interface.
>>>
>>> Best regards, Ilya Maximets.
>>>
>>>>
>>>> It seemed that Restarting the host's network service automatically 
>>>> triggered behavior: /as /usr/bin/ovs-vsctl del-port br-oam bond1/
>>>>
>>>> Also, I restart host which causes the same issue, would you please 
>>>> help check and give some advice, thx~
>>>>
>>>> //An

_______________________________________________
discuss mailing list
[email protected]
https://mail.openvswitch.org/mailman/listinfo/ovs-discuss

Reply via email to