[Pacemaker] Clone config question

Tue May 19 11:20:31 UTC 2009

Hi Andrew,

 thanks for the pointer. It turns out that running the 
Filesystem-monitor operation on the nfs server killed it since it too 
had a /data filesystem mounted, which then (mistakenly) was being 
unmounted causing the crash... Changing the mountpoint on the 
apachefs-resource fixed this. However, now I see the apachefs-service 
being tried to run TWICE on each apa-node! From my log:

lrmd[3101]: 2009/05/19_13:09:25 debug: on_msg_perform_op: add an 
operation operation start[11] on ocf::Filesystem::apachefs:0 for client 
3104, its parameters: directory=[/var/www] fstype=[nfs] 
device=[192.168.202.47:/data/export] CRM_meta_timeout=[20000] 
CRM_meta_clone_max=[2] crm_feature_set=[3.0.1] 
CRM_meta_globally_unique=[false] CRM_meta_clone=[0] 
CRM_meta_clone_node_max=[1]  to the operation list.
lrmd[3101]: 2009/05/19_13:09:25 info: rsc:apachefs:0: start
stonithd[3102]: 2009/05/19_13:09:25 debug: external_set_config: called.
stonithd[3102]: 2009/05/19_13:09:25 debug: external_get_confignames: called.
lrmd[3101]: 2009/05/19_13:09:25 debug: on_msg_add_rsc:client [3104] adds 
resource apachefs:1
crmd[3104]: 2009/05/19_13:09:25 info: do_lrm_rsc_op: Performing 
key=55:14:0:e38d2516-b13e-4c2f-b44d-86e574c7dcb7 op=apachefs:1_start_0 )
lrmd[3101]: 2009/05/19_13:09:25 debug: on_msg_perform_op:2356: copying 
parameters for rsc apachefs:1
lrmd[3101]: 2009/05/19_13:09:25 debug: on_msg_perform_op: add an 
operation operation start[12] on ocf::Filesystem::apachefs:1 for client 
3104, its parameters: directory=[/var/www] fstype=[nfs] 
device=[192.168.202.47:/data/export] CRM_meta_timeout=[20000] 
CRM_meta_clone_max=[2] crm_feature_set=[3.0.1] 
CRM_meta_globally_unique=[false] CRM_meta_clone=[1] 
CRM_meta_clone_node_max=[1]  to the operation list.
lrmd[3101]: 2009/05/19_13:09:25 info: rsc:apachefs:1: start

It tried modifying the apache clone line to 'clone apache apacheservice 
meta clone-max="2" clone-node-max="1"' (which shouldn't be necessary) 
but that didn;'t help either...

Should I maybe NOT use a clone for this and write out two seperate 
apache-configs for both nodes?

Andrew Beekhof wrote:
> On Tue, May 19, 2009 at 10:24 AM, Mark Schenk <M.M.A.Schenk at tudelft.nl> wrote:
>   
>> Hi,
>>
>>   I am trying to run cloned resources on part of my cluster, however I must
>> be mistaken on how to use the location-directive. This is the relevant part
>> of my config:
>>
>> node $id="8ce10d28-174c-4fc3-bac4-b2bf21ef6a38" apa1
>> node $id="09c41339-21db-4250-83b9-bac11a6c25de" apa2
>> node $id="ea3da999-829d-4042-a5d5-9ee0ad27f989" mysql1
>> node $id="23f84b10-ad8c-43a9-ab23-344d06254948" mysql2
>> node $id="47c29bf2-ac4a-4873-9b91-f05b8e41af2d" nfs1
>> node $id="e588fc86-92d8-4658-a394-3e031171bcd0" nfs2
>> primitive apachefs ocf:heartbeat:Filesystem \
>>         params device="192.168.202.47:/data/export" directory="/data"
>> fstype="nfs"
>> primitive apacheserver ocf:heartbeat:Dummy
>> group apacheservice apachefs apacheserver
>> clone apache apacheservice \
>>         meta clone-max="2"
>> location apaapa1 apache 100: apa1
>> location apaapa2 apache 100: apa2
>> property $id="cib-bootstrap-options" \
>>         dc-version="1.0.3-b133b3f19797c00f9189f4b66b513963f9d25db9" \
>>         symmetric-cluster="false" \
>>         last-lrm-refresh="1242719884"
>>
>>
>> What I thought this would do was: start (cloned) resources apachefs and
>> apacheserver only on nodes apa1 and apa2 (since symmetric-cluster=false and
>> having the two location directives). What I can see happening is that the
>> apachefs resource is tried on other nodes as well, most importantly also on
>> nfs1 and nfs2, which causes these to crash :-(.  Could someone point out how
>> I can get this working the way I want?
>>     
>
> It shouldn't be starting apache anywhere except apa1 & 2, but we do
> check _every_ node to see if its running (so that we can satisfy the
> constraints by stopping apache if we find it running somewhere it
> shouldn't be).
>
> _______________________________________________
> Pacemaker mailing list
> Pacemaker at oss.clusterlabs.org
> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
>   

-- 
_____________________________________________________________________
Mark Schenk
Project Manager
Shared Service Centre ICT - 3xO
Technische Universiteit Delft
M.M.A.Schenk at tudelft.nl
Room 3B-41
Postbus 354
2600 AJ Delft
Landbergstraat 15
2628 CE DELFT             ___________________________________________
+31 (0)15 27 85448        | Reality is that which, when you
www.tudelft.nl            | stop believing in it, doesn't go away.
_____________________________________________________________________

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.clusterlabs.org/pipermail/pacemaker/attachments/20090519/0ea61edf/attachment-0002.htm>