[zabbix-plugin] When stopping zabbix whole cluster goes down

[Sam Stoelinga]

Whole cluster goes down if you execute crm resource stop p_zabbix-server. This is because the resource vip__management (ocf::fuel:ns_IPaddr2) will also stop.

Steps to reproduce:
1. Deploy environment using Fuel 6.1 with Zabbix plugin enabled.
2. After deployment execute crm resource stop p_zabbix-server

Current result:
The vip_management resource also becomes stopped. Rendering lots of services unaccessible.

Expected result:
Ability to independently stop zabbix server wtihout influencing other services.

After some more research it looks like this is the line that's causing this behaviour:
cibadmin --query | grep vip_management | grep zabbix
<rsc_colocation id="vip_management-with-zabbix-server" rsc="vip__management" score="INFINITY" with-rsc="p_zabbix-server"/>

The current workaround is to remove that line from the cib with:
cibadmin --delete --xml-text '<rsc_colocation id="vip_management-with-zabbix-server"/>'

Recommended way of achieving this is through Pacemaker resource groups:
http://clusterlabs.org/doc/en-US/Pacemaker/1.0/html/Pacemaker_Explained/s-resource-sets-collocation.html

Example 6.15. A group resource with the equivalent colocation rules

  <group id="dummy">
    <primitive id="A" .../>
    <primitive id="B" .../>
    <primitive id="C" .../>
    <primitive id="D" .../>
  </group>

This notation can also be used in this context to tell the cluster that a set of resources must all be located with a common peer, but have no dependencies on each other. In this scenario, unlike the previous on, B would be allowed to remain active even if A or C (or both) were inactive.

[Oleksiy Molchanov (omolchanov)]

Zabbix team, can you check?

[Bartosz Kupidura (zynzel)]

This was intentional behavior. Zabbix deployed with zabbix-snmp-plugin required to be run on the same host as mgmt_vip.

Now we can reserve VIP address from plugin. We will create dedicated VIP for snmp-plugin and force zabbix to be runned on the same machine. Or we will use group resource.

We will remove collocation for zabbix+vip_mgmt.

[Sam Stoelinga (sammiestoel)]

You can still have the requirement for the services to be colocated on the same host. This bug is about the vip_management resource being stopped when zabbix-server is being stopped. You can still have them be colocated on the same host but use a resource group instead of a hard dependency of colocation. (That's how I understood the docs of pacemaker)

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/207011

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/207017

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/207017
Committed: https://git.openstack.org/cgit/stackforge/fuel-plugin-zabbix-snmptrapd/commit/?id=66e007426ceaacf0c1dc7fc5869264b8b8f6e112
Submitter: Jenkins
Branch: master

commit 66e007426ceaacf0c1dc7fc5869264b8b8f6e112
Author: Bartosz Kupidura <email address hidden>
Date: Wed Jul 29 14:18:57 2015 +0200

    Create group of resources for zabbix and mgmt vip

    Change-Id: I15bd9d65f3730eead4c4816f7655c6f89977a886
    Partial-Bug: #1478425

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/207011
Committed: https://git.openstack.org/cgit/stackforge/fuel-plugin-external-zabbix/commit/?id=d3bf8c5425d3a5addd3a48e71ecc4925b2ef755d
Submitter: Jenkins
Branch: master

commit d3bf8c5425d3a5addd3a48e71ecc4925b2ef755d
Author: Bartosz Kupidura <email address hidden>
Date: Wed Jul 29 13:52:46 2015 +0200

    Remove collocation between zabbix and mgmt vip

    Change-Id: I4acd08b1f75890dd64a3a283e3a3423cddedb37d
    Closes-Bug: #1478425

[Nikita Marchenko (nmarchenko)]

verified on 224 ISO

[Denis Meltsaykin (dmeltsaykin)]

The plugin repository has no branches 6.1/7.0/et cetera, moreover, packages that are being built from this repo are not a part of Fuel ISO and official Fuel packages repository. Closing this as Won't Fix.