JFYI............
I figure out that channel bonding doesn't work with "default" iptables
enabled. It has to be disabled, otherwise, I think DPM is being
firewalled out. Or, "bond0" (whatever the interface) needs to explicitly
defined in the iptables rules, which I yet to try.
Cheers,
Santanu
On 09/03/2011 15:48, Simon George wrote:
> Hi Santanu,
>
> we have the same bonding setup as you at RHUL and it works.
> i.e. nothing special on the switch,
> module options:
> options bond0 miimon=100 updelay=2000 downdelay=2000 mode=6
>
> Once we had a load of problems due to one of the ports used in the
> bond being faulty. But this looked like intermittent packet loss, not
> what you are seeing.
>
> Simon
>
> On 09/03/2011 14:28, Santanu Das wrote:
>> On 09/03/11 14:17, Sam Skipsey wrote:
>>> [ ... ]
>>> We do IEEE 802.3ad link aggregation (that's mode=4 ) here, which seems
>>> to work for everything, if you take care to use the right
>>> xmit_hash_policy.
>>
>> I have one machine with LACP/802.3ad link aggregation but that was for
>> virtual machine and that's the only way VMware ESXi works. I don't see
>> any difference between these two types, in terms or connecting nodes,
>> data transfer etc. Does anyone have a clue why it should be a problem
>> for DPM?
>>
>> Cheers,
>> Santanu
>>
>>
>>> Sam
>>>
>>>> Cheers,
>>>> Santanu
>>>>
>>>>
>>>>
>>>> On 09/03/11 13:16, Ewan MacMahon wrote:
>>>>
>>>> -----Original Message-----
>>>> From: GRIDPP2: Deployment and support of SRM and local storage
>>>> management
>>>> [mailto:[log in to unmask]] On Behalf Of Santanu Das
>>>>
>>>> There are the configuration files:
>>>>
>>>>
>>>> [root@disk09 network-scripts]# cat save.ifcfg-bond0
>>>> DEVICE=bond0
>>>> BOOTPROTO=dhcp
>>>> TYPE=bonding
>>>> ONBOOT=yes
>>>>
>>>> You don't appear to be setting BONDING_OPTS. Also, how have you
>>>> configured the switch?
>>>>
>>>> Ewan
>>>>
>>>>
|