[Cluster-devel] [PATCH] dlm_controld: add multi-home

[Cluster-devel] [PATCH] dlm_controld: add multi-home

[Patrick Caulfield]

This patch adds multi-home capability to dlm_controld. If a node has
more than one address then the DLM will be told about all of them,
rather than just the first as at present.

It is assumed that sctp will be used as a transport protocol in this
case, but dlm_controld does not force this. It might be more appropriate
to set this in conga, but I'm not sure.

Signed-Off-By: Patrick Caulfield <pcaulfie@redhat.com>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: mhome.patch
Type: text/x-patch
Size: 2170 bytes
Desc: not available
URL: <http://listman.redhat.com/archives/cluster-devel/attachments/20071210/f7780bb2/attachment.bin>

[Cluster-devel] [PATCH] dlm_controld: add multi-home

[Steven Whitehouse]

Hi,

On Mon, 2007-12-10 at 15:22 +0000, Patrick Caulfield wrote:
> This patch adds multi-home capability to dlm_controld. If a node has
> more than one address then the DLM will be told about all of them,
> rather than just the first as at present.
> 
> It is assumed that sctp will be used as a transport protocol in this
> case, but dlm_controld does not force this. It might be more appropriate
> to set this in conga, but I'm not sure.
> 
> Signed-Off-By: Patrick Caulfield <pcaulfie@redhat.com>

Are you sure that 4 addresses will be enough? seems a little low to me,
or did I misunderstand the purpose of MAX_NODE_ADDRESSES?

Steve.

[Cluster-devel] [PATCH] dlm_controld: add multi-home

[Fabio M. Di Nitto]

On Mon, 10 Dec 2007, Patrick Caulfield wrote:

> This patch adds multi-home capability to dlm_controld. If a node has
> more than one address then the DLM will be told about all of them,
> rather than just the first as at present.
>
> It is assumed that sctp will be used as a transport protocol in this
> case, but dlm_controld does not force this. It might be more appropriate
> to set this in conga, but I'm not sure.
>
> Signed-Off-By: Patrick Caulfield <pcaulfie@redhat.com>
>

+#define MAX_NODE_ADDRESSES 4

can we please make these kind of things fully dynamic?

I have 7 ip with 2 interfaces.. leave alone with VIP and other stuff... 
It would be an extra limitation to keep in mind when debugging why
some connections are working and others are not.

Thanks
Fabio

--
I'm going to make him an offer he can't refuse.

[Cluster-devel] [PATCH] dlm_controld: add multi-home

[Patrick Caulfield]

Steven Whitehouse wrote:
> Hi,
> 
> On Mon, 2007-12-10 at 15:22 +0000, Patrick Caulfield wrote:
>> This patch adds multi-home capability to dlm_controld. If a node has
>> more than one address then the DLM will be told about all of them,
>> rather than just the first as at present.
>>
>> It is assumed that sctp will be used as a transport protocol in this
>> case, but dlm_controld does not force this. It might be more appropriate
>> to set this in conga, but I'm not sure.
>>
>> Signed-Off-By: Patrick Caulfield <pcaulfie@redhat.com>
> 
> Are you sure that 4 addresses will be enough? seems a little low to me,
> or did I misunderstand the purpose of MAX_NODE_ADDRESSES?
> 


It's set to 3 in the kernel !

Patrick

[Cluster-devel] [PATCH] dlm_controld: add multi-home

[David Teigland]

On Mon, Dec 10, 2007 at 04:44:20PM +0100, Fabio M. Di Nitto wrote:
> On Mon, 10 Dec 2007, Patrick Caulfield wrote:
> 
> >This patch adds multi-home capability to dlm_controld. If a node has
> >more than one address then the DLM will be told about all of them,
> >rather than just the first as at present.
> >
> >It is assumed that sctp will be used as a transport protocol in this
> >case, but dlm_controld does not force this. It might be more appropriate
> >to set this in conga, but I'm not sure.
> >
> >Signed-Off-By: Patrick Caulfield <pcaulfie@redhat.com>
> >
> 
> +#define MAX_NODE_ADDRESSES 4
> 
> can we please make these kind of things fully dynamic?
> 
> I have 7 ip with 2 interfaces.. leave alone with VIP and other stuff... 
> It would be an extra limitation to keep in mind when debugging why
> some connections are working and others are not.

These addresses come from the openais multiring configuration.  The jump
from 1 to 2 openais rings is going to be challenging enough, IME.  As
Patrick points out, dlm-kernel supports up to 3.

How about this, once someone demonstrates a functional cluster running
just two rings and two sctp paths for the dlm, I'll bump up the kernel
limit.

[Cluster-devel] [PATCH] dlm_controld: add multi-home

[Joel Becker]

On Mon, Dec 10, 2007 at 09:59:35AM -0600, David Teigland wrote:
> These addresses come from the openais multiring configuration.  The jump
> from 1 to 2 openais rings is going to be challenging enough, IME.  As
> Patrick points out, dlm-kernel supports up to 3.

	Whatever happened to "we'll only run one cluster at a time"? :-)

Joel

-- 

"The nice thing about egotists is that they don't talk about other
 people."
         - Lucille S. Harper

Joel Becker
Principal Software Developer
Oracle
E-mail: joel.becker at oracle.com
Phone: (650) 506-8127

[Cluster-devel] [PATCH] dlm_controld: add multi-home

[David Teigland]

On Mon, Dec 10, 2007 at 12:48:49PM -0800, Joel Becker wrote:
> On Mon, Dec 10, 2007 at 09:59:35AM -0600, David Teigland wrote:
> > These addresses come from the openais multiring configuration.  The jump
> > from 1 to 2 openais rings is going to be challenging enough, IME.  As
> > Patrick points out, dlm-kernel supports up to 3.
> 
> 	Whatever happened to "we'll only run one cluster at a time"? :-)

Those are multiple rings among the same nodes for the sake of redundancy
in one cluster; it's some kind of variation of the totem protocol IIRC,
Steve Dake can explain it.  It's most definately not for multiple
clusters.

Dave