Cluster Server 8.0 Bundled Agents Reference Guide - Windows
- Introducing VCS bundled agents
- Storage agents
- Network agents
- FileShare agents
- Services and applications agents
- VCS infrastructure and support agents
- VCS hardware replication agent for NetApp
- Troubleshooting bundled agents
Service group behavior
Consider the following actions to better understand the service group behavior in this configuration.
The Online functions of all the resources in LocGrp (local service group) are called starting from the lower-most resource on local1. In this configuration, WebRes depends on RGRes. The RGRes Online function is called before the WebRes Online function. In the Online function, the RemoteGroup agent communicates to the remote cluster and authenticates the specified user. The RemoteGroup agent brings RemGrp online on remote1. RemGrp is in an ONLINE state on remote1 and RGRes is in an ONLINE state on local1. The Online function of WebRes is then called. WebRes is now in an ONLINE state.
Thus, you have established an application-level dependency across two different VCS clusters. The WebRes does not go online before the RGRes goes online. RGRes does not go online before the RemGrp goes online.
LocGrp is online on the local1. WebRes and RGRes are both online on local1. RemGrp is online on remote1. If RemGrp faults on remote1, RGRes detects that RemGrp has gone OFFLINE or has FAULTED on remote1 and goes into a FAULTED state. All the resources in LocGrp are taken offline on local1. LocGrp now fails over to local2. All the resources including WebRes and RGRes go into an ONLINE state on local2. As part of the fail over of LocGrp on local2, RemGrp goes online on remote2. LocGrp shows ONLINE on node local2 and OFFLINE OR FAULTED on local1.
LocGrp is online on local1 and RemGrp is online on remote1. If you take LocGrp offline, all the resources dependent on RGRes are taken offline. First, the WebRes resource goes offline and then the offline for RGRes is called. As the ControlMode is OnOff, the RemoteGroup agent communicates to the remote cluster and issues an offline command for RemGrp. Once RemGrp has gone offline, RGRes goes into OFFLINE state. Thus, the Web-server is taken offline before the database goes offline.