Table of Contents
Resource Group (Service) Manager
Running under CMAN (and using the DLM), rgmanager manages and provides failover capabilities for collections of resources called services, resource groups, or resource trees in a cluster. These resource groups are tree-structured, and have parent-child dependency and inheritance relationships within each subtree.
There are many resource types provided with rgmanager; the most important being the service resource type. The service resource type is more or less a container resource type. It has two primary purposes:
- provide a link between cluster.conf and rgmanager's placement policy
- provide ordering guarantees for children based on resource type
For a full list of resource types available with rgmanager, check here.
How rgmanager Works
- FailoverDomains - How rgmanager's failover domains work
- ServicePolicies - Rgmanager's service startup and recovery policies
- ResourceTrees - How rgmanager's resource trees work, including start/stop orders and inheritance
- ServiceOperationalBehaviors - How rgmanager's operations work and what states mean
- VirtualMachineBehaviors - Special things to remember when running VMs in a rgmanager cluster
- ResourceActions - What agent actions rgmanager uses and how to customize their behavior from cluster.conf
- EventScripting - If rgmanager's failover & recovery policies are not enough for you, write your own!
- OCF RA API Draft - How resource agents are (supposed to) work
- rgmanager mostly implements this, but does not correctly call monitor. Instead, it calls status.
How to use rgmanager
- HighlyAvailableNFS - Using NFS with rgmanager. Note: In Red Hat Enterprise Linux, only Method 1 is available.
- LVMFailover - Failover of a volume group without Clustered Logical Volume Management
- SambaFailover - Failover of a Samba share
- MultipleInstanceServices - If you want to run the same thing on many nodes, here's how.
- KvmMigration - How to configure ssh for use with KVM virtual machines managed by RGManager
- MultiHeadNFS - EXPERIMENTAL - Using NFS with Linux Virtual Server / Direct Routing to achieve a scalable, single-IP file server.
- PostgreSQLFailover - PostgreSQL 8
- Cluster Suite NFS Cookbook - a little bit out of date, but still lots of useful information.
Rgmanager may be configured for various log levels using different syslog facilities.
- Using cluster.conf:
<rm log_level="..." log_facility="..." ... >
- log_level is a number from 0..7, where 7 is 'debug' and 0 is 'emergency only'. The default value is 4.
- log_facility is a string containing a facility name, such as daemon, local4, or similar. Possible values for this are described in syslog.conf(5) manual page in the section entitled SELECTORS. The default value is daemon.
- Using the init script (which references /etc/sysconfig/cluster). Add the following to enable debug log level to /etc/sysconfig/cluster:
- This is the same as changing log_level to 7
It may be desirable to redirect rgmanager's logging messages to a different file (instead of /var/log/messages). In order to do this, place a line like the following in /etc/syslog.conf and restart syslog:
(The former assumes you have set log_facility to local4 in cluster.conf; the default is daemon. For more information, see the manual page for syslog.conf(5).
Event Scripting Configuration
RIND (is not dependencies) is talked about on the EventScripting page.
http://www.clusterlabs.org/wiki/Install - Installation guide for Pacemaker, a proposed replacement for rgmanager.