[ClusterLabs] Master/Slave DRBD not active on asymmetric cluster

[Bruyninckx Kristof]

Hello,

Currently I've tried to setup a 3 node asymmetric cluster, with the 3th node only being used as a tie breaker.

monnod01 & monnod02 :

*        centos 7.3

*        pacemaker-1.1.15-11.el7_3.2.x86_64,

*        corosync-2.4.0-4.el7.x86_64

*        drbd84-utils-8.9.8-1.el7.elrepo.x86_64

*        PostgreSQL 9.4
monquor:

*        centos 7.3

*        pacemaker-1.1.15-11.el7_3.2.x86_64

*        corosync-2.4.0-4.el7.x86_64

*        no drbd installed.

Now I've noticed that the master/slave drbd resource only activates the master side and not also the slave side allowing the drbd to actually sync between each other. I've setup a 2 node cluster, and there it works without any issue.
But when I try to do the same, but with a 3th node, and

pcs property set symmetric-cluster=false

For some reason it keeps adding the 3the node as a stopped resource in the master slave setup and it doesn't mention a slave resource.

pcs status
Online: [ monnod01 monnod02 monquor ]

Full list of resources:

ClusterIP      (ocf::heartbeat:IPaddr2):       Started monnod01
Master/Slave Set: drbd-demo-resource-clone [drbd-demo-resource]
     Masters: [ monnod01 ]
     Stopped: [ monquor ]

Resource created with the following

pcs -f drbd_cfg resource create drbd-demo-resource ocf:linbit:drbd drbd_resource=drbd-demo op monitor interval=10s
pcs -f drbd_cfg resource master drbd-demo-resource-clone drbd-demo-resource master-max=1 master-node-max=1 clone-max=2 clone-node-max=1 notify=true

Even though I've used location constraints on the master slave resource allowing it only access to the 2 nodes.

[root at monnod01 ~]# pcs constraint
  Resource: drbd-demo-resource-clone
    Enabled on: monnod01 (score:INFINITY)
    Enabled on: monnod02 (score:INFINITY)

The actual failover itself works, so it activates the DRBD disk, mounts it and starts up the db service which access the files on this drbd disk.
But since the slave drbd is never started, it'll never actually perform the drbd sync between the disks.
What am I missing to actually make the master/slave resource ignore the 3 node and startup the master and slave resource ?
Does DRBD need to be installed on the 3th node as well ?

I've put the complete output of the commands in the attachment of the mail.

Met vriendelijke groeten / Meilleures salutations / Best regards

Kristof Bruyninckx
System Engineer



-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.clusterlabs.org/pipermail/users/attachments/20170314/22aeded9/attachment-0001.html>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: output.txt
URL: <http://lists.clusterlabs.org/pipermail/users/attachments/20170314/22aeded9/attachment-0001.txt>