To enable incremental backups of virtual disks, Changed Block Tracking (CBT) must be used for the first full backup. (CBT is enabled for backups by default.)
Best Practices for SAN Transport
- SAN transport cannot be used by virtual machine proxies; use a HotAdd configuration to deploy a virtual proxy.
- SAN transport generally provides the best performance for thick disk restores, but the worst performance for restoring thin disks. For thin disk restores, NBD or NBDSSL is usually faster.
- For Enterprise and Datacenter editions of Windows Server 2008, the default SAN policy is offline. SAN policy should be set to onlineAll for Windows Server 2008 proxies, and SAN disks should be set to read-only except for restores. The diskpart utility can be used to clear the read-only flag.
Best Practices for HotAdd
- HotAdd relies on the SCSI protocol:
- With VDDK 6.5 or later, VMware recommends the paravirtual SCSI (PVSCSI) controller.
- For more information, see Virtual Disk Development Kit Release Notes.
- For older VDDK versions, use the LSI SCSI controller.
- HotAdd does not support IDE disks.
- Helper virtual machines are not required for HotAdd proxies using VADP.
- When deploying the Virtual Server Agent for HotAdd backups, choose the datastore with the largest VMFS block size to ensure backups can mount and back up virtual machines residing on all datastores. For VMFS-3, a proxy that needs to back up and restore very large virtual disks should be deployed on volumes that support a large block size, with the block size of the proxy datastore matching the block size of the datastore for the disk being backed up. VMFS-5 uses a consistent file block size and can handle volumes up to about 60TB.
- When HotAdd disks are backed up, a snapshot and redo log are created. While the HotAdd disk is still attached, do not remove the snapshot or the virtual machine being backed up. If the VM is removed while the disk is still attached, clean up of redo logs fails, and you must remove virtual disks from the backup appliance manually. If the snapshot is removed, the redo log could be left in an unconsolidated state.
- If you use the vSphere Client to remove all disks on a controller, the entry for the controller is also removed from vSphere.
- Virtual Windows disks created by HotAdd backup or restore operations might have different disk signatures than the original virtual disks.
- If the VM that is being backed up or restored and the VSA proxy for the operation are in different ESX clusters, the operation might fail.
Best Practices for NBDSSL
- In ESXi 5.0 and later, default NFC timeouts can be set in the VixDiskLib configuration file. If no timeout is specified, older versions of ESX or ESXi hold the corresponding disk open until vpxa or hostd is restarted. As a starting point for NBD and NBDSSL transport, set Accept and Request timeouts to 3 minutes, Read timeouts to 1 minute, Write timeouts to 10 minutes, and timeouts for nfcFssrvr and nfcFssrvrWrite to 0. You might need to lengthen timeouts on slow networks, especially for NBDSSL.
- A VMDK can fail to open if too many NFC connections are made to an ESX host. For more information, see VDDK library returns the error: Failed to open NBD extent, NBD_ERR_GENERIC (1022543).
Best Practices for IntelliSnap
- In environments leveraging Fibre Channel storage (required for HDS), install the Virtual Server Agent and MediaAgent on a physical computer.
- In array management. enter storage addresses for iSCSI and NFS in the same format used for ESX servers. For example, use an IP Address for both entries.
- When using NFS storage, enter each IP address that is used into array management. Entering only a single IP for a management interface is not sufficient.
- The Virtual Server Agent proxy must have access to the storage network. If you have an isolated network, an additional network connection must be added to the proxy.
- IntelliSnap performs a full backup. Switching to or from a IntelliSnap backup causes the next backup to be a full backup.
Last modified: 10/8/2019 4:07:07 PM