Mailing List Archive: Linux-HA: Users

Dependency on dm_multipath

Index | Next | Previous | View Threaded

cyruspy at gmail

Jul 6, 2008, 2:01 AM

Post #1 of 7 (205 views)
Permalink

Dependency on dm_multipath

Hi,

I would like to put xen domUs under control of hb, some of them will
depend on a ocfs2 filesystem (already under control of xen) and others
on dm_multipath devices (shared LUNs), is there any agent/method to
take care of this resources to add them as dependency to the relevant
domUs?

Regards,
Ciro
_______________________________________________
Linux-HA mailing list
Linux-HA[at]lists.linux-ha.org
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems

lmb at suse

Jul 7, 2008, 6:51 AM

Post #2 of 7 (186 views)
Permalink

Re: Dependency on dm_multipath [In reply to]

On 2008-07-06T05:01:18, Ciro Iriarte <cyruspy[at]gmail.com> wrote:

> Hi,
>
> I would like to put xen domUs under control of hb, some of them will
> depend on a ocfs2 filesystem (already under control of xen) and others
> on dm_multipath devices (shared LUNs), is there any agent/method to
> take care of this resources to add them as dependency to the relevant
> domUs?

I'm not quite sure what you mean by "under control of xen" (as Xen does
not control ocfs2) - why would you want to add dm_multipath as a
resource? They are available on all nodes already, there is nothing for
heartbeat to control?

Regards,
Lars

--
Teamlead Kernel, SuSE Labs, Research and Development
SUSE LINUX Products GmbH, GF: Markus Rex, HRB 16746 (AG Nürnberg)
"Experience is the name everyone gives to their mistakes." -- Oscar Wilde

_______________________________________________
Linux-HA mailing list
Linux-HA[at]lists.linux-ha.org
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems

cyruspy at gmail

Jul 8, 2008, 8:23 AM

Post #3 of 7 (182 views)
Permalink

Re: Dependency on dm_multipath [In reply to]

2008/7/7 Lars Marowsky-Bree <lmb[at]suse.de>:
> On 2008-07-06T05:01:18, Ciro Iriarte <cyruspy[at]gmail.com> wrote:
>
>> Hi,
>>
>> I would like to put xen domUs under control of hb, some of them will
>> depend on a ocfs2 filesystem (already under control of xen) and others
>> on dm_multipath devices (shared LUNs), is there any agent/method to
>> take care of this resources to add them as dependency to the relevant
>> domUs?
>
> I'm not quite sure what you mean by "under control of xen" (as Xen does
> not control ocfs2) - why would you want to add dm_multipath as a
> resource? They are available on all nodes already, there is nothing for
> heartbeat to control?
>
>
> Regards,
> Lars
>

Sorry, my mistake... Please read: "under control of heartbeat" (via
the Filesystem agent).... Now, I would like to make sure that I can
access the LUNs that are used for root FS in other domUs before
starting them..

Regards,
Ciro
_______________________________________________
Linux-HA mailing list
Linux-HA[at]lists.linux-ha.org
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems

mike.kershaw at gmail

Jul 8, 2008, 12:46 PM

Post #4 of 7 (180 views)
Permalink

Re: Dependency on dm_multipath [In reply to]

I don't know if there's anything built in, but I would imagine that
you could just create a contraint, to be sure that your resource
controlling your DomU's, aren't started before your ocfs2 resource. I
do the same to make sure that my ocfs resource starts after my evms
resource:

<rsc_order id="bcf-ocfs2_afterevms" from="bcf-ocfs2_cloneset" action="start"
to="evmscloneset" to_action="start" type="after"/>

Hope this kind of hits what you're looking for!

Mike

On Tue, Jul 8, 2008 at 11:23 AM, Ciro Iriarte <cyruspy[at]gmail.com> wrote:
> 2008/7/7 Lars Marowsky-Bree <lmb[at]suse.de>:
>> On 2008-07-06T05:01:18, Ciro Iriarte <cyruspy[at]gmail.com> wrote:
>>
>>> Hi,
>>>
>>> I would like to put xen domUs under control of hb, some of them will
>>> depend on a ocfs2 filesystem (already under control of xen) and others
>>> on dm_multipath devices (shared LUNs), is there any agent/method to
>>> take care of this resources to add them as dependency to the relevant
>>> domUs?
>>
>> I'm not quite sure what you mean by "under control of xen" (as Xen does
>> not control ocfs2) - why would you want to add dm_multipath as a
>> resource? They are available on all nodes already, there is nothing for
>> heartbeat to control?
>>
>>
>> Regards,
>> Lars
>>
>
> Sorry, my mistake... Please read: "under control of heartbeat" (via
> the Filesystem agent).... Now, I would like to make sure that I can
> access the LUNs that are used for root FS in other domUs before
> starting them..
>
>
> Regards,
> Ciro
> _______________________________________________
> Linux-HA mailing list
> Linux-HA[at]lists.linux-ha.org
> http://lists.linux-ha.org/mailman/listinfo/linux-ha
> See also: http://linux-ha.org/ReportingProblems
>
_______________________________________________
Linux-HA mailing list
Linux-HA[at]lists.linux-ha.org
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems

cyruspy at gmail

Jul 8, 2008, 3:49 PM

Post #5 of 7 (181 views)
Permalink

Re: Dependency on dm_multipath [In reply to]

2008/7/8 Michael Kershaw <mike.kershaw[at]gmail.com>:
> I don't know if there's anything built in, but I would imagine that
> you could just create a contraint, to be sure that your resource
> controlling your DomU's, aren't started before your ocfs2 resource. I
> do the same to make sure that my ocfs resource starts after my evms
> resource:
>
> <rsc_order id="bcf-ocfs2_afterevms" from="bcf-ocfs2_cloneset" action="start"
> to="evmscloneset" to_action="start" type="after"/>
>
>
> Hope this kind of hits what you're looking for!
>
> Mike
>
>
Hi,

Sorry, I wasn't clear again.

* domUs that use file stored on ocfs2 as disk images:
dom1, dom2, dom3

* domUs that use LUNs as system disks
domU4, domU5

For group1 I can create a dependency on the ocfs2 volume, that's clear.

For group2 I need to make sure that the LUNs are available, the issue
is how do I do that?

Regards,
CI.-
_______________________________________________
Linux-HA mailing list
Linux-HA[at]lists.linux-ha.org
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems

hunvagyok at freemail

Jul 8, 2008, 11:39 PM

Post #6 of 7 (178 views)
Permalink

Re: Dependency on dm_multipath [In reply to]

Hi,

On Tue, 2008-07-08 at 18:49 -0400, Ciro Iriarte wrote:
> For group2 I need to make sure that the LUNs are available, the issue
> is how do I do that?

I think you are approaching this problem from the wrong direction. You
should have 1 LUN which is managed by multipathd. Then you get evms on
top of the dm device to manage the LUN. It's an extra layer but gives
you flexibility, simplifies your SAN fabric and simplifies your
multipath config also.

I would create 1 evms volume for the ocfs2 image pool and several evms
volumes for domUs that need real block device.

domUs on file images can be dependant of ocfs2 resource
domUs on evms volumes can be dependant of evmsd perhaps EvmsSCC service.

I have the same setup and it works great. No need to bring multipath
into the HA, it's a horrible enough thing itself, just put evms on top.

Regards,
Ivan

_______________________________________________
Linux-HA mailing list
Linux-HA[at]lists.linux-ha.org
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems

lmb at suse

Jul 9, 2008, 5:29 AM

Post #7 of 7 (169 views)
Permalink

Re: Dependency on dm_multipath [In reply to]

On 2008-07-08T18:49:53, Ciro Iriarte <cyruspy[at]gmail.com> wrote:

> For group2 I need to make sure that the LUNs are available, the issue
> is how do I do that?

Well, two answers -

You could indeed write a pseudo-RA which monitors the DM-MPIO devs, or
feed their status directly into the CIB from multipathd (via an
unmanaged resource), or monitor the devs externally and modify a node
attribute accordingly (ie, "connected-to-san") and create a
rsc_location dependency. This is going to involve writing code.

Or you could simply assume that the likelihood of the DM-MPIO devs being
unavailable is really low (afterall, MPIO implies redundant connection);
if they really aren't available where the cluster tries to instantiate
the DomU, the start will simply fail and be retried on another node
within 1 second or so.

Personally, I would probably go with the second choice; lower
complexity.

Regards,
Lars

--
Teamlead Kernel, SuSE Labs, Research and Development
SUSE LINUX Products GmbH, GF: Markus Rex, HRB 16746 (AG Nürnberg)
"Experience is the name everyone gives to their mistakes." -- Oscar Wilde

_______________________________________________
Linux-HA mailing list
Linux-HA[at]lists.linux-ha.org
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems

Index | Next | Previous | View Threaded

Interested in having your list archived? Contact lists@gossamer-threads.com