Objective 3.4 – Configure and Manage VMFS Datastores
Identify VMFS file system attributes
First of all the VMFS must reside on a SCSI disk or volume; this can be presented by Fibre Channel, iSCSI or a local SAS controller. NOTE: whilst installing ESX on SATA is supported presenting a shared VMFS on said storage is not.
A VMFS volume can be a maximum of 64TB in size; this is achieved using extents; each extent or datastore can be a maximum of 2TB(-512 bytes).
The datastores are limited to 2TB(-512 bytes) because (i believe) the VMFS partition table uses the MBR partition format and this partitioning method limits the logical block addressing to 4bytes thus we only have 4 bytes to address all the sectors on the disk.
A VMFS cannot be deleted ot reformatted if one or more virtual machine locks exist; these locks are stored in the form fo metadata. On disk locks are updated using nonpersistent SCSI locks (SCSI reserve / reset commands).
VMFS-2 can be upgraded to VMFS-3.
Raw Device Maps use a VMDK pointer on the VMFS; the filesystem of the RDM is whatever is compatible with the virtual machine; the RDM doesn’t get encapsulated by the VMFS.
VMFS-3 is hierarchical and has block sizes of 1MB through 4MB for large files such as vmdk, vswp etc. ; the sub-block size is 64KB to handle small files such as vmx, nvram etc.
VMFS metadata have a .fs extension.
Determine the appropriate Datastore location/configuration for given virtual machines
To determine the appropriate location and configuration for a given virtual machine or set of virtual machines consider the following:
- Performance requirements
- High availabity requirements
- Point-in-time restoration
- Backup requirements
- Replication requirements
The budget will have a major impact on the storage array choice and ultimately functionality such as:
- disk and connectivity performance
- highly available storage processors and fabric
- SAN snapshot functionality
The performance requirements will in conjuction with budget be the major factor in deciding what storage array you use.
Determine use cases for multiple VMFS Datastores
Each VMFS should have a dedicated LUN and that LUN should be backed by a RAID array which meets the requirements of the said virtual machine or virtual machines.
When sizing the datastore consider the following:
- Space for growth
- Contention of the datastore
- Management overhead
- Virtual machine applications
A large datastore that allows growth would require close monitoring to ensure the LUN queue doesn’t become overly contended. Heavily used virtual machines should either have a dedicated datastore or have a low virtual machine to datastore ratio e.g. 3:1.
Virtual Machine placement methods:
The predictive method uses multiple datastores with different characteristics; placement of the virtual machine data will be determined by the virtual machine requirements and datastore characteristics.
The adaptive method uses datastores with the same characteristics; placement of the virtual machine data will be determined by the datastore performance threshold i.e. is there capacity available.
Create/Configure VMFS Datastores
VMFS datastores are created via the vSphere client.
Configuration tab > storage > Add storage > Disk / LUN > Next > select the LUN > Next > Next > Name the datastore > select the block size and capacity > Next > Finish.
Once you have added the VMFS datastore all the applicable ESX servers will automatically have access to the volume; previous versions required a storage rescan.
Attach existing Datastore to new ESX host
When a ESX host attaches an existing datastore it uses the exsiting disk signature. Datastores are added to the ESX host via the vSphere client.
Configuration tab > storage > Add storage > Disk / LUN > Next > Select the datastore with an existing label > Next > Keep existing signature > Next > Finish.
Manage VMFS Datastores
Datastores can be grouped for organisational purposes; to group datastores connect to vCenter via the vSphere client.
Inventory > Datastores > New folder > [drag and drop the datastores to the folder]
Datastores can be unmounted using the vSphere client. You can only unmount two datastore classes: NFS datastores and VMFS datastores which have been mounted without resignaturing.
Inventory > Datastores > [Right click the datastore unmount]
Datastores can be deleted using the vSphere client but depending on the datastore type depends on what actually happens; deleting a VMFS datastore destroys all the data, remember if any locks are in place you will not be able to delete the datastore. Deleting a NFS datastore just deletes the mount, the actual data is left intact.
Inventory > Datastores > [Right click the datastore delete]
Grow VMFS volumes
VMFS datastores can be grown to 2TB(-512 bytes) or the datastore can be extended above 2TB using extents.
A VMFS datastore can be grown using the vSphere client.
Configuration tab > Storage > select the storage > properties > select increase > select the volume > Next > Next > configure extent size > Next > Finish.
A VMFS datastore can be extended past 2TB by adding an extent. Extents are added in the same way as above; the only difference being you’re increasing the datastore by using another device / LUN.
When using extents note the first extent in the set contains the metadata thus loss or corruption of this extent will lead to data loss on all extents.