Database Reference
In-Depth Information
Hot CPU add
Increasing disk size
Utilization of vSphere snapshots
Pausing/resuming of a virtual machine
Memory overcommitment leading to virtual swapping or memory ballooning
Yes, for SQL Server AlwaysOn FCI, VMware does not support vMotion of the SQL
Server nodes. In addition, VMware does not support the use of the paravirtualized SCSI
adapter, and customers should follow the support and configuration guidance provided
in the SQL Server Clustering Guides for the version of vSphere running in your
environment. For more information on what is and is not supported by VMware,
bookmark
http://kb.vmware.com/kb/1037959
.
For a link to the Clustering Guide
specific to your version of vSphere, bookmark
http://kb.vmware.com/kb/1004617
.
From an operational perspective, it is important to set the automation level of all virtual
machines in an AlwaysOn FCI to Partially Automated. This will allow for the automatic
placement of the virtual machines during creation and will also serve to provide
migration recommendations, but it will not migrate the virtual machines. The vSphere
Clustering Guide for MSCS will contain more information on how to configure this
setting for your SQL Server AlwaysOn FCI virtual machines. We do want to stress that
vSphere HA is supported; however, it is important that SQL Server virtual machines
participating in the SQL Server AlwaysOn FCI never—and we mean
never
—reside on
the same host. There are several methods by which to configure this option, but we
recommend using vSphere DRS groups and VM-Host affinity rules. How to configure
this is detailed out in the Setup Guide for MSCS available for your version of vSphere
(use
http://kb.vmware.com/kb/1004617
to obtain the guide).
If you have an ESXi host with 10 or more RDMs used in the SQL Server AlwaysOn FCI
implementation, a reboot of the ESXi host may take a long time. Therefore, VMware
recommends marking the passive MSCS LUNs as perennially reserved. This will
reduce the time it takes the ESXi host to boot. For more information on how to configure
a perennially reserved LUN for your version of vSphere, KB 1016106
(
http://kb.vmware.com/kb/1016106
)
contains the answers.
The final item to discuss concerns increased operational coordination between the
vSphere team and the SQL Server team. Due to the fact SQL Servers participating in an
AlwaysOn FCI implementation cannot be vMotioned, any actions that require the
vSphere team to evacuate an ESXi host, such as applying an update, require the SQL
Server team to fail over the services from one SQL Server node to the other node. Some
customers view this as high operational overhead and as a reason not to virtualize SQL
Server. Other customers view this as a minor inconvenience to providing a more
resilient and reliable infrastructure for their customers.
