Книга: Mastering VMware® Infrastructure3

Creating and Managing VMFS Datastores

Creating and Managing VMFS Datastores

Microsoft has NTFS, Linux has EXT3, and so it is only fair that VMware have its own proprietary file system: VMFS. The VMware File System (VMFS) is a high-performance, distributed journaling file system used to house virtual machine files, ISO files, and templates in a VI3 environment. Any fibre channel, iSCSI, or local storage pool can be formatted as VMFS and used by an ESX Server host. Network storage located on NAS devices cannot be formatted as VMFS data-stores but still offer some of the same advantages. The VMFS found in the latest version of ESX Server is VMFS-3, which presents a significant upgrade from its predecessor VMFS-2.

VMFS protects against data integrity problems by writing all updates to a serial log on the disk before updating the original log. Postfailure, the server will restore the data to the prefailure state and recover any unsaved data by writing it to the intended prefailure location. Perhaps the most significant enhancement to VMFS-3 is its ability to have subdirectories, thus allowing virtual machine disk files to be located in respective folders under the VMFS volume parent label.

A VMFS volume stores all of the files needed by virtual machines, including:

? .vmx, the virtual machine configuration file.

? .vmx.lck, the virtual machine lock file created when a virtual machine is in a powered-on state.

? .nvram, the virtual machine BIOS.

? .vmdk, the virtual machine hard disk.

? .vmsd, the dictionary file for snapshots and the associated vmdk.

? .vmem, the virtual machine memory mapped to a file when the virtual machine is in a powered-on state.

? .vmss, the virtual machine suspend file created when a virtual machine is in a suspended state.

? -Snapshot#.vmsn, the virtual machine snapshot configuration.

? .vmtm, the configuration file for virtual machines in a team.

? -flat.vmdk, a pre-allocated disk file that holds virtual machine data.

? f001.vmdk, the first extent of pre-allocated disk files that hold virtual machine data split into 2GB files; additional files increment the numerical value.

? s001.vmdk, the first extent of expandable disk files that are split into 2GB files; additional files increment the numerical value.

? -delta.vmdk, the snapshot differences file.

VMFS is a clustered file system that allows multiple physical ESX Server hosts to simultaneously read and write to the same shared storage location. The block size of a VMFS volume can be configured as 1 MB, 2 MB, 4 MB, or 8 MB. Each of the block sizes corresponds to a maximum file size of 256GB, 512GB, 1024GB, and 2048GB, respectively.

To create a VMFS datastore on a fibre channel or iSCSI LUN using the VI Client, follow these steps: 

1. Use the VI Client to connect to a VirtualCenter or an ESX Server host.

2. Select the hostname in the inventory panel and then click the Configuration tab.

3. Select Storage (SCSI, SAN, and NFS) from the Hardware menu.

4. Click the Add Storage link.

5. As shown in Figure 4.36, select the radio button labeled Disk/LUN and then click Next.

Figure 4.36 The Disk/LUN option in the Add Storage wizard formats allows you to configure a fibre channel or iSCSI SAN LUN as a VMFS volume. 

6. As shown in Figure 4.37, select the appropriate SAN device for the VMFS datastore (i.e., vmhba1:0:21) and click Next.

Figure 4.37 The list of SAN LUNs will include only the non-VMFS LUNs as available candidates for the new VMFS volume.

7. Click Next on the Current Disk Layout page.

8. Type a name for the new datastore in the Datastore Name text box and then click Next.

9. As shown in Figure 4.38, select a maximum file size and block size from the Maximum File Size drop-down list.

Figure 4.38 VMFS volumes can be configured to support a greater block size, which translates into greater file size capacity.

10. (Optional) If desired, though it's not recommended, the new VMFS datastore can be a portion of the full SAN LUN. Deselect the Maximize Capacity checkbox and specify a new size.

11. Review the VMFS volume configuration and then click Finish. 

Although using the VI Client to create VMFS volumes through VirtualCenter is a GUI-oriented and simplified approach, it is also possible to create VMFS volumes from a command line using the fdisk and vmkfstools utilities. Perform these steps to create a VMFS volume from a command line: 

1. Log in to a console session or use an application like Putty.exe to establish an SSH session to an ESX Server host.

2. Type the following command at the # prompt:

esxcfg-vmhbadevs

3. Determine if a valid partition table exists for the respective LUN by typing the following command at the # prompt:

fdisk -l /dev/sd?

where ? is the letter for the respective LUN. For example, /dev/sdf is shown in Figure 4.39.

Figure 4.39 The esxcfg-vmhbadevs command identifies all the available LUNs for an ESX Server host; the fdisk -l /dev/sd? command will identify a valid partition table on a LUN.

4. To create a new partition on a LUN, type the following at a # prompt:

fdisk /dev/sd?

where ? is the letter for the respective LUN.

Figure 4.40 To create a VMFS volume from the command line, you must create a partition on the target LUN.

5. Type n to add a new partition.

6. Type p to create the partition as a primary partition.

7. Type 1 for a partition number of 1.

8. Press the Enter key twice to accept the default values for the first and last cylinders.

9. Type p to view the partition configuration. Steps 5 through 9 are displayed in Figure 4.40. 

10. Once you've created the partition, define the partition type. At the Command (m for help) prompt, press the T key to enter the partition type.

11. At the Hex code prompt (type L to list codes), type fb to select the unknown code that corresponds to VMFS.

12. Type w to save the configuration changes. Figure 4.41 shows steps 10 through 12.

Figure 4.41 Once the partition is created, adjust the partition type to reflect the VMFS volume that will be created.

13. As shown in Figure 4.42, type the following command at the # prompt:

vmkfstools -C vmfs3 -S <VMFSNAME> vmhbaw:x:y:z

where <VMFSNAME> is the label to provide the VMFS volume, w is the HBA to use, x is the target ID for the storage processor, y is the LUN ID, and z is the partition.

Figure 4.42 After the LUN is configured, use vmkfstools to create the VMFS volume.

Once the VMFS volume is created, the VI Client provides an easy means of managing the various properties of the VMFS volume. The LUN properties offer the ability to add extents as well as to change the datastore name, the active path, and the path selection policy. 

Adding extents to a datastore lets you increase the size of an extent greater than the 2TB limit. This does not allow file sizes to exceed the 2TB limit but only the size of the VMFS volume. Be careful when working with adding extents because the LUN that is added as an extent is wiped of all its data during the process. This can result in unintentional data loss. LUNs that are available as extent candidates are those that are not already formatted with VMFS, leaving empty or NTFS-formatted LUNS as viable candidates. (In other words, ESX will not eat its own!)

While adding an extent through the VI Client or command line is an easy way to provide more space, it is a better practice to manage VMFS volume size from the storage device. To add an extent to an existing datastore, perform these steps: 

1. Use the VI Client to connect to a VirtualCenter or an ESX Server host.

2. Select the hostname from the inventory pane and then click the Configuration tab.

3. Select Storage (SCSI, SAN, and NFS) from the Hardware menu. 

4. Select the datastore to which the extent will be added.

5. Click the Properties link.

6. Click the Add Extent button, shown in Figure 4.43, on the datastore properties.

Figure 4.43 The properties page of a datastore lets you add an extent to increase the available storage space a datastore offers.

7. Choose an extent candidate on the Extent Device page of the Add Extent wizard, shown in Figure 4.44, and then click Next.

Figure 4.44 When adding an extent, be sure the selected extent candidate does not currently hold any critical data. All data is removed from an extent candidate when added to a datastore.

8. Click Next on the Current Disk Layout page.

9. (Optional) Although it is not recommended, the Maximum Capacity checkbox can be deselected and a custom value can be provided for the amount of space to use from the LUN.

10. Click Next on the Extent Size page.

11. Review the settings for the new extent and then click Finish.

12. Identify the new extent displayed in the Extents list and the new Total Formatted Capacity of the datastore, and then click Close. 

13. As shown in Figure 4.45, identify the additional extent in the details pane for the datastore.

Figure 4.45 All extents for a datastore are reflected in the Total Formatted Capacity for the datastore and can be seen in the Extents section of the datastore details. 

Once an extent is complete, the datastore will maintain the relationship until the datastore is removed from the host computer. An individual extent cannot be removed from a datastore.

All the financial, human, and time investment put into building a solid virtual infrastructure would be for nothing if ESX Server did not offer a way of ensuring accessing to VMFS volumes in the face of hardware failure. ESX Server has a native multipathing feature that allows for redundant access to VMFS volumes across an available path. 

It doesn't seem likely that critical production systems with large financial investments in network storage would be left susceptible to single points of failure — which is why, in most cases, a storage infrastructure built to host critical data is done with redundancy at each level of the deployment. For example, a solid fibre channel infrastructure would include multiple HBAs per ESX host, connected to multiple fibre channel switches, connected to multiple storage processors on the storage device. In a situation where an ESX host has two HBAs and a storage device has two storage processors, there are four (2?2) different paths that can be assembled to provide access to LUNs. This concept, called multipathing, involves the use of redundant storage components to ensure consistent and reliable transfer of data. Figure 4.46 depicts a fibre channel infrastructure with redundant components at each level, which provides for exactly four distinct paths for accessing the available LUNs. 

The multipathing capability built into ESX Server offers two different methods for ensuring consistent access: the most recently used (MRU) and the fixed path. As shown in Figure 4.47, the details section of an ESX datastore will identify the current path selection policy as well as the total number of available paths. The default policy, MRU, provides failover when a device is not functional but does not failback to the original device when the device is repaired. As the name suggests, an ESX host configured with an MRU policy will continue to transfer data across the most recently used path until that path is no longer available or is manually adjusted.

Figure 4.46 ESX Server has a native multipathing capability that allows for continued access to storage data in the event of hardware failure. With two HBAs and two storage processors, there exist exactly four paths that can be used to reach LUNs on a fibre channel storage device.

 Figure 4.47 The Details of a datastore configured on a fibre channel LUN identifies the current path selection policy.

The second policy, fixed path, requires administrators to define the order of preference for the available paths. The fixed path policy, like the MRU, provides failover in the event that hardware fails, but it also provides failback upon availability of any preferred path as defined.

Perform the following steps to edit the path selection policy for a datastore: 

1. Use the VI Client to connect to a VirtualCenter or an ESX Server host.

2. Select the hostname in the inventory panel and then click the Configuration tab.

3. Select Storage (SCSI, SAN, and NFS) from the Hardware menu.

4. Select a datastore and review the details section.

5. Click the Properties link for the selected datastore.

6. Click the Manage Paths button in the Datastore properties box.

7. Click the Change button in the Policy section of the properties box.

8. As shown in Figure 4.48, select the Fixed radio button.

Figure 4.48 You can edit the path selection policy on a per-LUN basis. 

9. Click OK. 

10. Click OK.

11. Click Close.

Perform the following steps to change the active path for a LUN:

1. Use the VI Client to connect to a VirtualCenter or an ESX Server host.

2. Select the hostname in the inventory panel and then click the Configuration tab.

3. Select Storage (SCSI, SAN, and NFS) from the Hardware menu.

4. Select a datastore and review the details section.

5. Click the Properties link for the selected datastore.

6. Click the Manage Paths button in the Datastore properties box, shown in Figure 4.49.

Figure 4.49 The Manage Paths detail box identifies the active and standby paths for a LUN and can be used to manually select a new active path.

7. Select the existing Active path and then click the Change button beneath the list of available paths.

8. Click the Disabled radio button to force the path to change to a different available path, shown in Figure 4.50.

Figure 4.50 Disabling the active path of a LUN forces a new active path. 

9. Repeat the process until the desired path is shown as the Active path. 

Regardless of the LUN design strategy or multipathing policy put in place, virtual infrastructure administrators should take a very active approach to monitoring virtual machines to ensure that their strategies continue to maintain adequate performance levels.

Оглавление книги