Alan,

Thanks for your quick reply with such details. I got it.

Lizhong

Alan Maguire wrote:
> Lizhong Li wrote:
>> Renee and Anurag,
>>
>> I have two questions,:
>> one is how to judge whether a link should be plumbed by nwamd, i.e., 
>> it should
>> be cable-pluged, but how to decide it by command like 'dladm' ? So 
>> the test case
>> can check if the output from 'nwamadm list' is right.
>>   
> Unfortunately the ability of a link NCU to display
> state reflecting whether a cable is plugged in depends on whether the
> associated driver supports link notifications (what
> you've probably seen us refer to as DL_NOTE_LINK_UP/DOWN).
> Without such notifications, NWAM simply assumes that
> there is a cable plugged in. If the underlying driver
> supports DL_NOTE_LINK_UP/DOWN, the link
> NCU will reflect the plugged/unplugged state.
>
> IP NCUs will stay offline* forever in your case
> because they are all part of the same priority group,
> and one member of that group (nge0) succeeded in
> getting an IP address. The others are waiting for
> DHCP addresses. We do time out on these DHCP
> requests (if you run "nwamadm list -x" the auxiliary
> states of these offline* IP NCUs should show
> "timed out"), but the timeout will simply cause us
> to reassess the current priority group. Since the
> requirements of that priority group are met thanks to
> nge0, we don't switch (which would send all the link
> and interface NCUs offline). The idea of this policy is
> that we don't wait for wired links to come up forever if
> there's also a wireless link. If no wired links come up,
> we switch to wireless. In your case, you only have
> wired links, and one has come up. All the wired links are in the
> same shared priority group. Shared priority groups mean
> that if any link is active, the group is considered
> valid.
>
> On the surface, it sounds reasonable to send IP NCUs
> offline when the fail to get an address in a timely manner,
> but if I remember correctly, we are still retrying so
> theoretically an address could arrive.
>> The other question is if the ncu goes into the 'offline*' state, how 
>> long should
>> it change?  since I see it's in 'offline*' state as for ever, does it 
>> make sense ?
>>
>>   
> In this case, yes. If, for example, you had a wired and
> a wireless device, and the wired device wasn't plugged
> in, you'd see a switch to wireless eventually. Since
> this system only has wired devices, there's no other
> priority group to switch to, and the requirements
> of the group (one link being up) are also met.
>
> Alan

-- 
Thanks,
Lizhong


Reply via email to