Brocade Communications Systems StorageWorks 4400 - Enterprise Virtual Array Manual de usuario descargar pdf (Pagina 19)

Site A access to storage failure

When shared storage resources being used by SQL Server become unavailable, a cluster failover

event is triggered. Once the cluster node running SQL Server realizes the disk resources are

unavailable, Microsoft failover cluster attempts to move the SQL Server instance to the second node in

the list of preferred cluster nodes. If the second node in the list is the same site as the node previously

running site, the HP CLX EVA resource detect that the local array is no longer available. In most

cases, the desired behavior is to automatically fail over to Site B.

HP CLX EVA can only consistently, automatically fail over if the parameter Use non-current data OK is

set to YES in the Cluster Extension EVA resource property.

The Parameter Use non-current data OK specifies whether or not the EVA Cluster Extension can fail

over to the destination site in cases where the data at the destination site might not be current. This

might occur in the unlikely event that an HP Continuous Access EVA link goes down prior to the fail

over to the remote site:

o Setting this value to YES means that regardless of whether or not the data is current in the

remote site, the resource comes online.

o Setting the value to NO means that if the data is current in the remote site, the resource comes

online even though Use non-current data OK is set to NO. If the data is not current in the

remote site, or if the array cannot determine if data is current, the resource will not come

online.

Once SQL Server successfully fails over and is online in Site B, transactions can resume. At this point,

the data replication group goes into logging state, because there is no replication communication

between the two arrays. All subsequent I/Os are copied to the data replication group write history

log until the log becomes full, or until the replication link between the arrays is restored with

resynchronization before the log becomes full. If the write history log becomes full, the data

replication group is marked to do a full copy from Site B to Site A. If connectivity between the two

arrays is restored before the data replication group is marked for full copy, a delta (merge)

resynchronization occurs rather than a full copy.

Site A site failure

In the event of a full-site failure, the behavior is the same as in the Site A access to storage failure,

with the exception that the cluster will immediately try to bring the SQL Server instance online in Site

B. Automatic recovery of the clustered SQL Server instance once again relies on the Cluster Extension

EVA parameter Use non-current data OK for Microsoft Failover Cluster to successfully fail over the

SQL Server instance to Site B.

Failover during DR group resynchronization

CLX EVA has a resource property that determines Failover/Failback behavior that occurs when a site

failover occurs during resynchronization.

If the Failover/Failback property of the HP CLX EVA resource is set to RESYNCWAIT, HP EVA waits

until the resynchronization completes before successfully coming online. To ensure automatic failover,

RESYNCWAIT should be set, even though it is not the default value. FASTFAILBACK is only useful if

there is connectivity between the two arrays, which may not be the case for all disaster scenarios.

Figure 11 shows the location of the HP Cluster Extension resource property page under the

Parameters tab.

1 2 ... 14 15 16 17 18 19 20 21 22 23

Sin comentarios

Brocade Communications Systems StorageWorks 4400 - Enterprise Virtual Array Manual de usuario Pagina 19