Monitoring and controlling GeoCluster jobs

In a GeoCluster configuration, where data is stored on volumes local to each node and replicated to each node in the cluster, the GeoCluster Replicated Disk resource keeps the data synchronized between nodes of the cluster. You should also use the standard Windows cluster tools to monitor the status of the resource, but you should also use the following information to help you monitor the GeoCluster Replication Disk resource.

Do not use the Initiate Failure feature of Cluster Administrator to test failover of GeoCluster resources. Use other test methods, such as manually moving the group or unplugging the owning nodes network cable.

Do not use the Automatic Failback feature of Cluster Administrator. If you need to return ownership to the original node, wait until GeoCluster has completed mirroring from the new owning node back to the original owning node and then manually move the group.

If you change an IP address on any node of the cluster, you must stop and restart the cluster service on all of the nodes in the cluster in order for GeoCluster to detect the new IP address.

If you must reboot the owning node, you should move all of your cluster groups to another node before the reboot. This will ensure data integrity and allow you to make sure the applications come online before the node is rebooted.

Resolving an online pending GeoCluster Replicated Disk resource

When the GeoCluster Replicated Disk resource is in an online pending state, you are protected from possible data corruption. If you are using Windows 2003, review the description of the GeoCluster Replicated Disk Status resource to see why the GeoCluster Replicated Disk resource is in the online pending state. If you are using Windows 2008, you can see the online pending status directly in the description of the GeoCluster Replicated Disk resource. If the pending state were bypassed, the node where you are trying to bring the resource online would have incomplete data, which would then be replicated to the other nodes in the cluster. This state safeguards you from corrupting your data.

There are different options for resolving an online pending state, depending on whether your operating system supports snapshots. Therefore, some of the following options may not be displayed or may be disabled if they are not valid for your configuration.

If you are using Windows 2003, right-click on the online pending resource and select the desired control. The controls are described in the following tables.

If you are using Windows 2008, right-click the online pending resource, select Properties, select the Online Pending tab, and click the desired control. The controls are described in the following tables.

Windows 2003 Menu

Revert to snapshot

Windows 2008 Menu

Revert Snapshot

Description

If you have a snapshot of the target data available, you can revert to that data. If you revert to a snapshot, any data changes made after the snapshot’s specified date and time will be lost. A Double-Take Availability connection will be established to replicate the node’s data (at the snapshot point-in-time) to the other nodes.

Windows 2003 Menu

Discard target queue

Windows 2008 Menu

Discard Queue

Description

If you have data in the target queue, you can discard that data. If you discard the queued data, you will lose the changes associated with that data made on the previously owning node. A Double-Take Availability connection will be established to replicate the node’s data (without the data that was in queue) to the other nodes.

Windows 2003 Menu

Force Resource Offline

Windows 2008 Menu

Fail Resource

Description

If you are using Windows 2003, you can force the resource offline. If you are using Windows 2008, you can fail the resource. In either case, no Double-Take Availabilityconnection will be established.

Windows 2003 Menu

Verify Group

Windows 2008 Menu

Verify Group

Description

With this option and snapshot capability, you can test the data on the node before deciding whether to use it. If you select this option, a snapshot of the node’s current Double-Take Availability data will be taken, but the GeoCluster Replicated Disk resource does not come online, allowing you to check the data. (This means there is no Double-Take Availability connection established at this time.) Once the snapshot is taken, you can test the data on the node to see if it is viable. Once you have tested the data, you need to right-click on the online pending resource again and accept or reject the data.

Windows 2003 Menu

Accept Data

Windows 2008 Menu

Description

If you accept the data, the current data on the node will be used, and a Double-Take Availability connection will be established to replicate the current node’s data to the other nodes. If any other nodes in the cluster contain more recent data, this node will overwrite that data and it will be lost.

Windows 2003 Menu

Reject Data

Windows 2008 Menu

Reject

Description

If you reject the data, the node will be reverted to the snapshot that was taken when you selected the Verify Group option. Any changes made on the node after that snapshot was created will be lost. This option essentially takes you back to where you were, allowing you the opportunity to check other nodes for more recent data.

If you have multiple GeoCluster Replicated Disks in the same group and have selected to reject the data after verifying the group, the rejection processing may take several minutes.

GeoCluster Replicated Disk Status Resource

The function of the GeoCluster Replicated Disk Status resource (also displayed as GRD Status) varies between Windows 2003 and Windows 2008. In both operating systems, it is automatically created when the first GeoCluster Replicated Disk resource is created in a group. Once the status resource is created, it will exist as long as there is a GeoCluster Replicated Disk resource in the group. When the last GeoCluster Replicated Disk resource in a group is deleted, the status resource will be deleted. Only one status resource is created per group.

If you are using Windows 2003, the description of the status resource corresponds to various states of your Double-Take Availability data. By reviewing the status descriptions, you can tell at-a-glance the state of your Double-Take Availability data. If you are using Windows 2008, these status descriptions are seen directly in the GeoCluster Replicated Disk resource description, rather than the GeoCluster Replicated Disk Status resource.

For example, you may see the status "The status of all targets is OK." This indicates the data on each target node is in a good state. Another message may be "Target target_name is queuing. Data in queue on target." This indicates the data on the specified target is not up-to-date. Because there is data in queue on the target, that has not been written to disk yet, the target data is out-of-date. Or you may see either of the following status descriptions.

Target target_name is pending. Data integrity not guaranteed.
Target target_name is suspect. Data integrity not guaranteed.

These messages indicate the data on the specified target node is not in a good state. This could be because a mirror is in progress, an operation has been dropped on the target, or another Double-Take Availabilityprocessing issue. Check the Double-Take Availability logs for more information. As long as the status is pending, data integrity cannot be guaranteed on the specified target node.

The text of the descriptions may vary between Windows 2003 and Windows 2008.

Another function of the status resource, for both Windows 2003 and Windows 2008, is to keep you from moving the GeoCluster Replicated Disk resource to another node at the wrong time and potentially corrupting your data. If the GeoCluster Replicated Disk resource was moved while the status resource is in a pending or queuing state, the new node would have incomplete data, which would then be replicated to the other nodes in the cluster. This resource safeguards you from corrupting your data.