Re: [opencontrail-dev] service instance health-check bug

2016-09-19 Thread 钱行 





I use vyatta as firewallservice and version of service chain is v1.




On 09/19/2016 14:18，

Ganesha H V

wrote：








What is the image and version of Service Chain you are using? 






Thanks & Regards,
Ganesha HV.
"To be conscious that you're ignorant, is a great step to success" - Dr.Kalam




From: Dev  on behalf of 钱行 
Sent: Monday, September 19, 2016 11:45:00 AM
To: dev
Subject: [opencontrail-dev] service instance health-check bug
 



hi,all


    I started a service-instance with 2 max-instances in active-active ha_mode. I call them vmA and vmB
    As usual, each of these instances has three interfaces, which are manage, left and right interface,with flow from left to right,
and I bonded health-check instance with left interfaces and right interfaces.
    If I shutdown the left interface in the vmA, the flows all go to vmB of course.
    BUT, if I shut down the right interface in the vmA, some flow still go to vmA, and the service-instance doesn't work well.
    So, we lose half of the flows when one of the right interfaces goes down.


    How can I solve it? Or, we should have a data struct like "risk group", which contains left and right interface peers, and once
one of the interface health-checked failed, all the interface in the group will be set inactive.




    Any help will be appreciated!
---
Hang Qian







___
Dev mailing list
Dev@lists.opencontrail.org
http://lists.opencontrail.org/mailman/listinfo/dev_lists.opencontrail.org

Re: [opencontrail-dev] service instance health-check bug

2016-09-19 Thread Ganesha H V 
What is the image and version of Service Chain you are using?



Thanks & Regards,
Ganesha HV.
"To be conscious that you're ignorant, is a great step to success" - Dr.Kalam

From: Dev  on behalf of 钱行 

Sent: Monday, September 19, 2016 11:45:00 AM
To: dev
Subject: [opencontrail-dev] service instance health-check bug

hi,all

I started a service-instance with 2 max-instances in active-active ha_mode. 
I call them vmA and vmB
As usual, each of these instances has three interfaces, which are manage, 
left and right interface,with flow from left to right,
and I bonded health-check instance with left interfaces and right interfaces.
If I shutdown the left interface in the vmA, the flows all go to vmB of 
course.
BUT, if I shut down the right interface in the vmA, some flow still go to 
vmA, and the service-instance doesn't work well.
So, we lose half of the flows when one of the right interfaces goes down.

How can I solve it? Or, we should have a data struct like "risk group", 
which contains left and right interface peers, and once
one of the interface health-checked failed, all the interface in the group will 
be set inactive.


Any help will be appreciated!
---
Hang Qian
___
Dev mailing list
Dev@lists.opencontrail.org
http://lists.opencontrail.org/mailman/listinfo/dev_lists.opencontrail.org