[ 
https://issues.apache.org/jira/browse/CLOUDSTACK-4405?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13752513#comment-13752513
 ] 

Kishan Kavala commented on CLOUDSTACK-4405:
-------------------------------------------

I was able to repo this issue and identified the root cause as follows:
1. Source host has bridge name in old format : cloudVirBr-<vlanId>
2. Destination host has no bridge for the vlan 
3. During prepare for migration, bridge is created on the destination host in 
new format br<physicalInterface>-<vlanId> (since there is no bridge created for 
the particular vlan)
4. Migration fails since there is no bridge in old format on the destination 
host

Possible Solutions:
1. Check the bridge format on source host and create the bridge in same format 
(cloudvirbr or breth) on destination host. This could result in a host having 2 
bridges for the same vlan. One in old format other in new format, and Vms in 
same vlan will also be split across these 2 bridges. 

2. Change all bridge names to new format. This would require all nics of guest 
Vms to be unplugged. This wont be acceptable in production environment.

3. Introduce a param (something like use.new.bridge.format) in agent.properties 
to specify which bridge format to use. By default new bridge format will be 
used. When set to false old bridge format will be used.

Option # 3 will address most of the issues. Using old bridge format always will 
solve migration issue.


                
> (Upgrade) Migrate failed between existing hosts and new hosts
> -------------------------------------------------------------
>
>                 Key: CLOUDSTACK-4405
>                 URL: https://issues.apache.org/jira/browse/CLOUDSTACK-4405
>             Project: CloudStack
>          Issue Type: Bug
>      Security Level: Public(Anyone can view this level - this is the 
> default.) 
>    Affects Versions: 4.1.0, 4.2.0
>         Environment: CS 4.1
>            Reporter: Wei Zhou
>            Assignee: Kishan Kavala
>            Priority: Blocker
>             Fix For: 4.1.1, 4.2.0, 4.2.1
>
>
> There are two hosts (cs-kvm001, cs-kvm002) in old 2.2.14 environment .
> After upgrade from 2.2.14 to 4.1, I added two new hosts (cs-kvm003, 
> cs-kvm004).
> The migration between cs-kvm001 and cs-kvm002, or cs-kvm003 and cs-kvm004 
> succeed.
> However, the migration from cs-kvm001/002 to the new hosts (cs-kvm003, 
> cs-kvm004) failed.
> 2013-08-19 16:57:31,051 DEBUG [kvm.resource.BridgeVifDriver] 
> (agentRequest-Handler-1:null) nic=[Nic:Guest-10.11.110.231-vlan://110]
> 2013-08-19 16:57:31,051 DEBUG [kvm.resource.BridgeVifDriver] 
> (agentRequest-Handler-1:null) creating a vlan dev and bridge for guest 
> traffic per traffic label cloudbr0
> 2013-08-19 16:57:31,051 DEBUG [utils.script.Script] 
> (agentRequest-Handler-1:null) Executing: /bin/bash -c brctl show | grep 
> cloudVirBr110
> 2013-08-19 16:57:31,063 DEBUG [utils.script.Script] 
> (agentRequest-Handler-1:null) Exit value is 1
> 2013-08-19 16:57:31,063 DEBUG [utils.script.Script] 
> (agentRequest-Handler-1:null)
> 2013-08-19 16:57:31,063 DEBUG [kvm.resource.BridgeVifDriver] 
> (agentRequest-Handler-1:null) Executing: 
> /usr/share/cloudstack-common/scripts/vm/network/vnet/modifyvlan.sh -v 110 -p 
> em1 -b brem1-110 -o add
> 2013-08-19 16:57:31,121 DEBUG [kvm.resource.BridgeVifDriver] 
> (agentRequest-Handler-1:null) Execution is successful.
> 2013-08-19 16:57:31,122 DEBUG [kvm.resource.BridgeVifDriver] 
> (agentRequest-Handler-1:null) Set name-type for VLAN subsystem. Should be 
> visible in /proc/net/vlan/config
> This is because the bridge name on old hosts are cloudVirBr110, and brem1-110 
> on new hosts.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

Reply via email to