Re: Corosync IPC over TCP

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Shridhar Sahukar napsal(a):
> 
> On 01/11/2013 11:45 AM, Vladimir Voznesensky wrote:
>> Why do you need a centralized corosync? It's not intended for
>> centralized usage, it's really ok to use it de-centralized.
> In the typical use case, the corosync daemon running on each node
> represents the existence of that node in the cluster, as it joins the
> totem ring. In our case, you could view the corosync daemon representing
> the entire chassis of 12 blades and not the individual blade. So the
> corosync cluster in our case is not a cluster of nodes, but a "cluster
> of group of nodes". That is why we want to run a single daemon per
> chassis and not a daemon on each node.
> 

I know some of group messaging systems are implementing this feature,
but actually, I'm not very sure about such feature. Actually, you are
probably first person asking for such feature, so my answer is "no, we
will not implement it, at least not in the near feature".

Of course, if you will decide to implement by yourself, I will be more
then happy to review/merge your patches and help with corosync internals.

Honza

> Regards,
> Shridhar
> 
>> 11.01.2013 19:44, Shridhar Sahukar пишет:
>>> Hi,
>>>
>>> We are trying to use corosync to bridge communication across two or
>>> more ATCA chassis. We do not want to run the corosync daemon on all
>>> the blades of the ATCA chassis, but we would need the applications
>>> running on all the cards to be able to talk to the centralized
>>> corosync deamon -- for example to use the cpg service.
>>>
>>> The existing architecture of corosync does not seem to support the
>>> above mentioned use case as it assumes that the corosync daemon is
>>> running on all the cards and the client and server are located on the
>>> same node/card. This restriction arises mainly due to use of unix
>>> domain sockets.
>>>
>>> Can we achieve the above use case if the underlying IPC is changed to
>>> TCP instead of unix sockets? What other issues do we need to address
>>> to have this working?
>>>
>>> Regards,
>>> Shridhar
>>> _______________________________________________
>>> discuss mailing list
>>> discuss@xxxxxxxxxxxx
>>> http://lists.corosync.org/mailman/listinfo/discuss
>>
> 
> _______________________________________________
> discuss mailing list
> discuss@xxxxxxxxxxxx
> http://lists.corosync.org/mailman/listinfo/discuss

_______________________________________________
discuss mailing list
discuss@xxxxxxxxxxxx
http://lists.corosync.org/mailman/listinfo/discuss



[Index of Archives]     [Linux Clusters]     [Corosync Project]     [Linux USB Devel]     [Linux Audio Users]     [Photo]     [Yosemite News]    [Yosemite Photos]    [Linux Kernel]     [Linux SCSI]     [X.Org]

  Powered by Linux