Open Stack

On 03/13/2015 05:54 AM, Georgios Dimitrakakis wrote:
>> On 03/12/2015 06:24 PM, Georgios Dimitrakakis wrote:
>>> Hello!
>>>
>>> I am facing a rather weird issue and would like to know if anyone has
>>> faced it before and the way you have solved it (if any).
>>>
>>> Firstly, I 'd like to let you know that I am running Icehouse with Nova-Legacy
>>> Networking.
>>>
>>> I have an application that launches and terminates VMs very frequently and
>>> automatically assigns to them
>>> floating IP addresses.
>>>
>>> As a result if an IP has been used once then the second time that is being used
>>> cannot be reached for at
>>> least 20minutes (this is the time frame if the newly launched VM gets the same
>>> IP address from one that
>>> has just been terminated). Furthermore, this issue only happens when the second
>>> VM is launched on a
>>> different physical node (hypervisor) than the first one.
>>>
>>> I believe that I have identified the problem to be ARP Caching on the router
>>> side since if I do:
>>>
>>> arping -U -I {Interface-Name} {IP-Address}
>>>
>>> in order to update the ARP Caches of the neighbors immediately the new VM can be
>>> reached on its floating
>>> IP address.
>>>
>>>
>>> Is this the proper way of solving it?
>>>
>>> Can I somehow "instruct" OpenStack not to use the first available IP address
>>> everytime but on a random order
>>> or the one that hasn't been used for enough time?
>>>
>>> Any other thoughts?
>>
>>
>> Just a thought, have you tried enabling "send_arp_for_ha" in nova.conf?
>>
>> Chris
>>
>
>
> Hi Chris! Thx for the feedback!
>
> send_arp_for_ha = True
>
> is already enabled in all nodes (controller and compute).
>
> Can I somehow verify that it's working?

You should be able to verify with a packet sniffer somewhere on your network.

> How often is it sending the ARP request? Does it send it every time nova configuration is changing?

I'm very much *not* an expert in this area, so I'm afraid I'm not the right 
person to ask about that.


Chris