On Wed, Nov 24, 2010 at 3:07 PM, Vit Pelcak <vpel...@suse.cz> wrote:
> Dne 23.11.2010 16:34, Andrew Beekhof napsal(a):
>>
>> On Tue, Nov 23, 2010 at 1:19 PM, Vit Pelcak<vpel...@suse.cz> wrote:
>>>
>>> Hello.
>>>
>>> I have prepared several scripts for automation of HA testing.
>>>
>>> I'd like to ask for some help with porting them into CTS, or at least
>>> part
>>> of them. As there is no documentation for CTS and I'm just learning
>>> Python,
>>> I could use some help here and there while trying to figure out how to do
>>> that.
>>>
>>> If I understand it right, CIB.py contains configuration of resources and
>>> new
>>> ones are added like:
>>> stonith_sbd_resource_template = """
>>> <resources>
>>> <primitive class="stonith" id="sbd_stonith" type="external/sbd">
>>> <meta_attributes id="sbd_stonith-meta_attributes">
>>> <nvpair id="sbd_stonith-meta_attributes-target-role" name="target-role"
>>> value="Started"/>
>>> </meta_attributes>
>>> <operations>
>>> <op id="sbd_stonith-monitor-1" name="monitor" interval="120s"
>>> prereq="nothing" timeout="300s"/>
>>> <op id="sbd_stonith-monitor-2" name="start" prereq="nothing"
>>> timeout="120s"/>
>>> <op id="sbd_stonith-monitor-3" name="stop" timeout="120s"/>
>>> </operations>
>>> </operations>
>>> <instance_attributes id="sbd_stonith-instance_attributes">
>>> <nvpair id="sbd_stonith-instance_attributes-sbd_device" name="sbd_device"
>>> value="%s"/>
>>> </instance_attributes>
>>> </primitive>
>>> </resources>"""
>>>
>>> Could this be replaced just by running crm configure .... ? I think, that
>>> is
>>> would be better (but I can be wrong, of course).
>>
>> Yes, we do that actually.
>> The above xml is only for older versions of pacemaker when the shell
>> didnt exist.
>
> Ah. I see:
>
> class HASI(CIB10):
> def add_resources(self):
> # DLM resource
> self._create('''primitive dlm ocf:pacemaker:controld op monitor
> interval=120s''')
> self._create('''clone dlm-clone dlm meta globally-unique=false
> interleave=true''')
>
> # O2CB resource
> self._create('''primitive o2cb ocf:ocfs2:o2cb op monitor
> interval=120s''')
> self._create('''clone o2cb-clone o2cb meta globally-unique=false
> interleave=true''')
> self._create('''colocation o2cb-with-dlm INFINITY: o2cb-clone
> dlm-clone''')
> self._create('''order start-o2cb-after-dlm mandatory: dlm-clone
> o2cb-clone''')
>
> BTW, shouldn't section for O2CB resource (and perhaps ohers) be updated also
> with recommended start/stop timeouts? Those default are perhaps not optimal.
> At least I see crm tool complaining about it when creating them.
Right, its not used much, it was mostly still in the PoC phase.
>
>>> Then I would call something like:
>>> stonith_resource = self.stonith_sbd_resource_template % (p_value)
>>>
>>> Is that right?
>>
>> yep
>>
>>> In the process of SBD creation, I need to call:
>>> sbd -d $sbd_disk create
>>>
>>> on one of nodes, and
>>>
>>> sbd -d $sbd_disk allocate $(hostname)
>>>
>>> on every node. How can I do that?
>>
>> CTS doesn't try to configure an entire cluster from scratch, there are
>> still some things that are specific to your test environment and this
>> looks like one of them (not all clusters have shared storage for
>> example)
>
> OK. Then I'll handle this thing by my own scripts which are not viable to be
> ported to CTS.
>
>> There is scope for creating more complex test scenarios though, see
>> the HAE section of the same file.
>
> I didn't find it.
You did :-)
> Did you mean this?:
Correct
> class HASI(CIB10):
> def add_resources(self):
> # DLM resource
> self._create('''primitive dlm ocf:pacemaker:controld op monitor
> interval=120s''')
> self._create('''clone dlm-clone dlm meta globally-unique=false
> interleave=true''')
>
> # O2CB resource
> self._create('''primitive o2cb ocf:ocfs2:o2cb op monitor
> interval=120s''')
> self._create('''clone o2cb-clone o2cb meta globally-unique=false
> interleave=true''')
> self._create('''colocation o2cb-with-dlm INFINITY: o2cb-clone
> dlm-clone''')
> self._create('''order start-o2cb-after-dlm mandatory: dlm-clone
> o2cb-clone''')
>
>> In that scenario one can assume shared storage so adding sbd wouldn't
>> be a problem.
>
> # SBD resource
> crm configure primitive sbd_stonith stonith:external/sbd meta
> target-role="Started" op monitor interval="15" timeout="15" params
> sbd_device="%s"
>
> ?
Possibly, never configured it myself.
>
>>> I assume, that CTS can handle formatting of disks for correct fs easily
>>> by
>>> running mkfs or any other commands just as bash would. But how can I make
>>> it
>>> do so on one specific machine while other machines would do something
>>> else?
>>
>> Not sure. It wasn't really designed for this.
>
> OK. Maybe some parts are not good idea to be ported to CTS.
>
>> Possibly you could just loop through the configured machines and treat
>> the first as special.
>
> I'll consider that, however, most probably, I'll prepare cluster by own
> scripts for specific scenario as well and then run CTS.
>
>
>> _______________________________________________
>> Pacemaker mailing list: Pacemaker@oss.clusterlabs.org
>> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
>>
>> Project Home: http://www.clusterlabs.org
>> Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
>> Bugs:
>> http://developerbugs.linux-foundation.org/enter_bug.cgi?product=Pacemaker
>
>
> _______________________________________________
> Pacemaker mailing list: Pacemaker@oss.clusterlabs.org
> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
>
> Project Home: http://www.clusterlabs.org
> Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
> Bugs:
> http://developerbugs.linux-foundation.org/enter_bug.cgi?product=Pacemaker
>
_______________________________________________
Pacemaker mailing list: Pacemaker@oss.clusterlabs.org
http://oss.clusterlabs.org/mailman/listinfo/pacemaker
Project Home: http://www.clusterlabs.org
Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
Bugs: http://developerbugs.linux-foundation.org/enter_bug.cgi?product=Pacemaker
