In large data centers, whenever the storage administrator requires to perform maintenance operations on a SAN infrastructure component, it’s difficult to identify all the entities in the data center are currently using it. Typically, the storage administrator needs to run various set of inquiries and provide list of transport identifiers (WWN mappings, iSCSI target IDs) that would be affected by the maintenance operation to the vSphere administrator who then, in turn, needs to search through the environment to identify the set of affected ESX hosts and the running virtual machines. This process is time consuming, tiring and error prone. Additionally due to the dynamic nature of VMware, the configuration information can change very quickly; the location of a virtual machine may be different than it was mere minutes ago.
VxDMP helps ease this painful process by providing visibility into the vSphere entities that are currently using a given storage infrastructure component along with the amount of I/O activity and connectivity issues that need to be remediated via maintenance operations. VxDMP also provides mechanisms to gracefully stop the usage of the storage infrastructure components, without any disruption in service to the virtual machines, enabling storage maintenance operations be carried out.
In the following sections, each of the types of storage maintenance operations will be examined in greater detail.
Storage Array Port/Controller maintenance: When the storage array port or controller needs to be quiesced for maintenance, the storage administrator simply provides the vSphere administrator the storage array name and the IDs of one or more ports (of a given storage controller) that are planned for maintenance. The vSphere administrator then uses the VxDMP tab within the data center view to enable or disable usage of the array ports from all VxDMP-controlled ESX hosts.
The vSphere administrator can quickly determine if the disable operation would lead to complete loss of connectivity of the storage from a host. If so, appropriate steps can be taken to either migrate the virtual machines running on the host to an alternate ESX host or inform the application administrator that the operation requires planned downtime.
SAN switch maintenance: When a SAN switch requires maintenance, the SAN administrator informs the vSphere administrator about the affected switch port mapping (initiator port and target array port). The vSphere administrator can then go to the VxDMP tab for the affected hosts and can selectively disable or enable the I/O activity on the specific initiator target transport links.
If the disable operation could cause complete loss of connectivity to a set of LUNs being used by the host, then the vSphere administrator is informed about it. If, after confirming that there is no active I/O on the affected LUNs, then the operation can be re-initiated specifying the force option.
This operation may also be done from the array target port view by disabling the selected HBA initiator ports that connect to the given SAN switch.
HBA maintenance: When a HBA on a host needs to be brought down for maintenance, the vSphere administrator can perform this operation without any disruption to virtual machines running on the host by disabling the I/O activity on access paths seen via the HBA.
This operation is similar to selecting each LUN individually and disabling paths that correspond to a given HBA but it much more convenient, since it can be performed once affecting all connections to the HBA.
So we can see how VxDMP makes the job of isolating storage infrastructure components for maintenance so easy. I would like to hear back from you as to how these operations helped cut down the time required to complete for such activities.