Re: [Openais] Multycast & unicast as fall back

2011-07-27 Thread Florian Haas
On 2011-07-27 10:43, Lars Marowsky-Bree wrote:
> On 2011-07-21T07:59:23, Steven Dake  wrote:
> 
>> There is no fallback.  You can specify one transport or the other.
>> Thinking a moment how to implement this type of feature, it could not be
>> reasonably implemented.
> 
> But the transport is per ring, is it not?

The most recent information available to me was that all rings have to
use the same transport. If that has changed, I'd be thrilled to hear
about it.

Florian



signature.asc
Description: OpenPGP digital signature
___
Openais mailing list
Openais@lists.linux-foundation.org
https://lists.linux-foundation.org/mailman/listinfo/openais

Re: [Openais] Multycast & unicast as fall back

2011-07-27 Thread Lars Marowsky-Bree
On 2011-07-21T07:59:23, Steven Dake  wrote:

> There is no fallback.  You can specify one transport or the other.
> Thinking a moment how to implement this type of feature, it could not be
> reasonably implemented.

But the transport is per ring, is it not?

Regards,
Lars

-- 
Architect Storage/HA, OPS Engineering, Novell, Inc.
SUSE LINUX Products GmbH, GF: Jeff Hawn, Jennifer Guild, Felix Imendörffer, HRB 
21284 (AG Nürnberg)
"Experience is the name everyone gives to their mistakes." -- Oscar Wilde

___
Openais mailing list
Openais@lists.linux-foundation.org
https://lists.linux-foundation.org/mailman/listinfo/openais


Re: [Openais] Multycast & unicast as fall back

2011-07-21 Thread Proskurin Kirill
On 07/21/2011 06:59 PM, Steven Dake wrote:

> There is no fallback.  You can specify one transport or the other.
> Thinking a moment how to implement this type of feature, it could not be
> reasonably implemented.

Well this is sad but ok, no problem.

> What type of app are you running on top of corosync?  The advantages of
> multicast is automatic growth (you don't have to know the node addresses
> ahead of time) and more throughput with less cpu utilization on high cpg
> message throughput.  The disadvantage is multicast is generally poorly
> implemented by switch vendors.

Yes, automatic grow is a killer feature but I just think how can we 
avoid network problems and unicast here is more redundant then 
multycast(in case of some network work - we have a really big and 
complex network). So I just make "non auto fallback" by make second 
config and switch to it if I need.

Thank for your work any way!


-- 
Best regards,
Proskurin Kirill
___
Openais mailing list
Openais@lists.linux-foundation.org
https://lists.linux-foundation.org/mailman/listinfo/openais


Re: [Openais] Multycast & unicast as fall back

2011-07-21 Thread Steven Dake
On 07/21/2011 06:27 AM, Proskurin Kirill wrote:
> On 07/21/2011 05:11 PM, Steven Dake wrote:
>> On 07/21/2011 02:30 AM, Proskurin Kirill wrote:
>>> Hello all.
>>>
>>> Is this possible to use multycast as primary way to communication in
>>> cluster but fall back to unicast transports if multycast is fail?
>>> Different rings with different transports?
>>>
>>> We have some problems in network switches and multycast just stop
>>> working and I start to think about this feature.
> 
>> Just use udpu entirely.  This feature is supported n 1.3.2+.
> 
> I`m on 1.4.0 now but I not wish to use unicast as production base - only
> if some problems with multycast occur.
> 
> 

There is no fallback.  You can specify one transport or the other.
Thinking a moment how to implement this type of feature, it could not be
reasonably implemented.

What type of app are you running on top of corosync?  The advantages of
multicast is automatic growth (you don't have to know the node addresses
ahead of time) and more throughput with less cpu utilization on high cpg
message throughput.  The disadvantage is multicast is generally poorly
implemented by switch vendors.

Regards
-steve
___
Openais mailing list
Openais@lists.linux-foundation.org
https://lists.linux-foundation.org/mailman/listinfo/openais


Re: [Openais] Multycast & unicast as fall back

2011-07-21 Thread Proskurin Kirill
On 07/21/2011 05:11 PM, Steven Dake wrote:
> On 07/21/2011 02:30 AM, Proskurin Kirill wrote:
>> Hello all.
>>
>> Is this possible to use multycast as primary way to communication in
>> cluster but fall back to unicast transports if multycast is fail?
>> Different rings with different transports?
>>
>> We have some problems in network switches and multycast just stop
>> working and I start to think about this feature.

> Just use udpu entirely.  This feature is supported n 1.3.2+.

I`m on 1.4.0 now but I not wish to use unicast as production base - only 
if some problems with multycast occur.


-- 
Best regards,
Proskurin Kirill
___
Openais mailing list
Openais@lists.linux-foundation.org
https://lists.linux-foundation.org/mailman/listinfo/openais


Re: [Openais] Multycast & unicast as fall back

2011-07-21 Thread Dan Frincu
Hi,

On 07/21/2011 12:30 PM, Proskurin Kirill wrote:
> Hello all.
>
> Is this possible to use multycast as primary way to communication in
> cluster but fall back to unicast transports if multycast is fail?
> Different rings with different transports?
>
> We have some problems in network switches and multycast just stop
> working and I start to think about this feature.
>
>
 From what I know (although I may be mistaken) it's a matter of 
either/or, so either multicast or unicast.

Have you tried looking into => 
http://www.corosync.org/doku.php?id=faq:cisco_switches maybe that helps.

Regards.
Dan
___
Openais mailing list
Openais@lists.linux-foundation.org
https://lists.linux-foundation.org/mailman/listinfo/openais


Re: [Openais] Multycast & unicast as fall back

2011-07-21 Thread Steven Dake
On 07/21/2011 02:30 AM, Proskurin Kirill wrote:
> Hello all.
> 
> Is this possible to use multycast as primary way to communication in 
> cluster but fall back to unicast transports if multycast is fail? 
> Different rings with different transports?
> 
> We have some problems in network switches and multycast just stop 
> working and I start to think about this feature.
> 
> 

Just use udpu entirely.  This feature is supported n 1.3.2+.

Regards
-steve
___
Openais mailing list
Openais@lists.linux-foundation.org
https://lists.linux-foundation.org/mailman/listinfo/openais


[Openais] Multycast & unicast as fall back

2011-07-21 Thread Proskurin Kirill
Hello all.

Is this possible to use multycast as primary way to communication in 
cluster but fall back to unicast transports if multycast is fail? 
Different rings with different transports?

We have some problems in network switches and multycast just stop 
working and I start to think about this feature.


-- 
Best regards,
Proskurin Kirill
___
Openais mailing list
Openais@lists.linux-foundation.org
https://lists.linux-foundation.org/mailman/listinfo/openais