[Pacemaker] Master/Slave Confusion

Tue Feb 2 04:10:00 UTC 2010

OK - it seems I've achieved what I want via the following configuration:

node testvm1
node testvm2
node testvm3
primitive LDAP lsb:ldap \
         op monitor interval="40s" \
         op monitor interval="41s" role="Master"
primitive LDAP-IP ocf:heartbeat:IPaddr2 \
         params ip="10.1.1.80" cidr_netmask="16" \
         op monitor interval="30s"
ms LDAP-clone LDAP \
         meta master-max="1" master-node-max="1" clone-max="2" 
clone-node-max="1"
location LDAP-IP-master LDAP-IP 10: testvm3
location LDAP-IP-slave LDAP-IP 5: testvm2
location LDAP-master LDAP-clone 10: testvm3
location LDAP-slave LDAP-clone 5: testvm2
colocation LDAP-with-IP inf: LDAP-IP LDAP-clone
order LDAP-after-IP inf: LDAP-IP:start LDAP-clone:promote
property $id="cib-bootstrap-options" \
         dc-version="1.0.5-462f1569a43740667daf7b0f6b521742e9eb8fa7" \
         cluster-infrastructure="openais" \
         expected-quorum-votes="3" \
         stonith-enabled="false" \
         symmetric-cluster="false"

testvm3 is the preferred server for LDAP and the floating IP, if testvm3 
goes down the IP floats over to testvm2.  However, when I do a status I 
see this:

============
Last updated: Mon Feb  1 19:51:34 2010
Stack: openais
Current DC: testvm1 - partition with quorum
Version: 1.0.5-462f1569a43740667daf7b0f6b521742e9eb8fa7
3 Nodes configured, 3 expected votes
2 Resources configured.
============

Online: [ testvm3 testvm2 testvm1 ]

LDAP-IP (ocf::heartbeat:IPaddr2):       Started testvm3
Master/Slave Set: LDAP-clone
         Slaves: [ testvm3 testvm2 ]

Both testvm2 and testvm3 are still "Slaves".  Does anyone know what this 
means?  Why isn't one of them a Master?  I'd rather configure the IP to 
follow the "Master", then what I'm doing now, which is simply tell the 
IP and LDAP that they prefer testvm3.  I have LDAP-clone and LDAP-IP 
colocated, but since the clone can exist on either node it doesn't mean 
much.  I think the IP is following the preference I set, and not 
'following the master' since there is no master?

How does one promote a slave to master automatically?

Thanks again for any insight!

-erich

Erich Weiler wrote:
>> if you make the LDAP daemon listen on all available interfaces, it 
>> will accept connections on the on-demand activated floating-ip.
> 
> Well, I'm trying to get this to work and running into a wall...  I've 
> got 3 servers, I want LDAP to run on testvm2 and testvm3.  I've 
> configured LDAP on those 2 servers.  Then I configured crm from scratch 
> as thus:
> 
> configure property stonith-enabled=falseconfigure primitive LDAP-IP 
> ocf:heartbeat:IPaddr2 params ip="10.1.1.80" cidr_netmask="16" op monitor 
> interval="30s"
> configure primitive LDAP lsb:ldap op monitor interval="40s"configure ms 
> LDAP-clone LDAP meta master-max="1" master-node-max="1" clone-max="2" 
> clone-node-max="1"
> configure colocation LDAP-with-IP inf: LDAP-IP LDAP-clone:Master
> configure order LDAP-after-IP inf: LDAP-IP:start LDAP-clone:promote
> 
> Here's the crm config at the moment:
> 
> node testvm1
> node testvm2
> node testvm3
> primitive LDAP lsb:ldap \
>         op monitor interval="40s"
> primitive LDAP-IP ocf:heartbeat:IPaddr2 \
>         params ip="10.1.1.79" cidr_netmask="16" \
>         op monitor interval="30s"
> ms LDAP-clone LDAP \
>         meta master-max="1" master-node-max="1" clone-max="2" 
> clone-node-max="1"
> colocation LDAP-with-IP inf: LDAP-IP LDAP-clone:Master
> order LDAP-after-IP inf: LDAP-IP:start LDAP-clone:promote
> property $id="cib-bootstrap-options" \
>         dc-version="1.0.5-462f1569a43740667daf7b0f6b521742e9eb8fa7" \
>         cluster-infrastructure="openais" \
>         expected-quorum-votes="3" \
>         stonith-enabled="false"
> 
> But when I check status, I see the IP is not started anywhere, LDAP is 
> started on testvm2 and testvm3 (like I wanted), but they are both listed 
> as slaves:
> 
> ============
> Last updated: Mon Feb  1 14:29:47 2010
> Stack: openais
> Current DC: testvm3 - partition with quorum
> Version: 1.0.5-462f1569a43740667daf7b0f6b521742e9eb8fa7
> 3 Nodes configured, 3 expected votes
> 2 Resources configured.
> ============
> 
> Online: [ testvm3 testvm2 testvm1 ]
> 
> Master/Slave Set: LDAP-clone
>         Slaves: [ testvm2 testvm3 ]
> 
> Failed actions:
>     LDAP_monitor_0 (node=(null), call=4, rc=5, status=complete): not 
> installed
>     LDAP:1_monitor_0 (node=(null), call=5, rc=5, status=complete): not 
> installed
> [root at testvm3 ~]#
> 
> So I guess my 3 questions are:
> 
> 1: Why isn't one of the LDAP servers being promoted to master?
> 2: Is the floating IP down because I've specified that it stick to the 
> master, but there is no master?
> 3: I'd like the LDAP master to exist on testvm3 (with the floating IP) 
> at all times, only failing to testvm2 if testvm3 goes down, but I'm not 
> clear on how to specify a preference for a node to be a master.
> 
> The "Failed actions" exist, I'm guessing, because I didn't install an 
> ldap server on testvm1, but I don't care about that because I only want 
> LDAP to stay on testvm2 and testvm3.
> 
> I feel like I'm close but just lack a little understanding.  The docs 
> have me almost there but I'm still a bit blurry obviously.  Any help 
> much appreciated!!  I think pacemaker will do great things for us if I 
> can get it working as expected....
> 
> _______________________________________________
> Pacemaker mailing list
> Pacemaker at oss.clusterlabs.org
> http://oss.clusterlabs.org/mailman/listinfo/pacemaker