[Sheepdog] PATCH S003: Handle master crashing before sending JOIN request

[Liu Yuan]

On 04/28/2012 05:49 PM, Shevek wrote:

> On Sat, 2012-04-28 at 17:34 +0800, Liu Yuan wrote:
>> On 04/28/2012 05:22 PM, Shevek wrote:
>>
>>>>> -               if (join_finished)
>>>>> +               if (join_finished || is_master(&cevent->sender))
>>> I believe this patch does NOT work:
>>>
>>> 1) is_master returns -1 for not-master, 0 means the given node is
>>> master, so this is apparently a misuse of is_master.
>>>
>>> However, even worse:
>>>
>>> 2) If we never got a join response, then line 388:
>>>     388 if (!cevent->blocked && cpg_node_equal(&cevent->sender,
>>> &this_node)) {
>>> never triggered, therefore we have never set join_finished, and
>>> nr_cpg_nodes is always 0 (look for every place where it got set), so
>>> is_master() is always 0.
>>>
>>> So, by inspection, this patch can never work, for both reasons.
>>>
>>
>>
>> Oh, yes. Thanks for pointing it out.
>>
>> But what if set set
>>
>>                 if (join_finished || (is_master(&cevent->sender >= 0)))
>>                         done = __corosync_dispatch_one(cevent);
>>                 else
>>                         done = !cevent->blocked;
>>
>> Then, the master will be called for sure, and would give other nodes a
>> join response.
> 
> As I said before under (2), is_master() will always return 0, since the
> only assignments to nr_cpg_nodes are within __corosync_dispatch_one()
> (which has never yet been called) or when first_node == 1 (which never
> happened).
> 
> So this new patch is still invalid.
> 


confchg event will be broadcast to all the nodes.

for our case, if is_master() returns >=0, it means the very node *is*
the master, then will call into __corosync_dispatch_one[1], where then
it will call into send_message(response).

so your assumption for (2) isn't valid.

[1]
if (join_finished || (is_master(&cevent->sender >= 0)))
	done = __corosync_dispatch_one(cevent);

Thanks,
Yuan