Accessing VMFS datastore
For data recovery purpose it's required to provide direct disk access to the investigated disks. Such conditions can be achieved by:
- connecting VMFS disk as local hard drive,
- connecting remotely via iSCSI,
- connecting remotely via SSH,
- recovering data from datastore with another Windows VM
Please note that ESX recognizes only VMFS partitions and we are unable to save to VMFS disks directly. Thus recovered data should be either transferred via network or saved to local hard drive with further upload via standard vSphere. This transfers are normally time consuming because of large amount of recovered data and slower network transfer speed. You need to take this operation into account when estimating recovery time.
VMFS drive as local disk in dedicated Windows PC
This is the most recommended method for data recovery. We recommend to connect all HDDs with datastores to dedicated Windows PC in order to gain access to VMFS partitions. This is the most robust way to obtain direct disk access and it has the highest speed data transfer.
RAID disks, iSCSI storage and NAS devices with VMFS partitions
If there is no issue with RAID disks or controller, array can be also moved to dedicated Win PC to process recovery. If iSCSI or NAS have no errors and you are sure there is no issue with them they could be connected\configured to allow access to VMFS datastore for Windows OS where recovery process is executed.
However, if you have any doubts that RAID\iSCSI\NAS integrity is compromised, you need to either connect HDDs\SSDs to Windows PC directly as standalone single drives or create disk images of these drives, also as individual disks, not as an image of RAID storage and mount these images to VMFS Recovery.
Main advantage of iSCSI interface over SSH is a speed. In our tests SSH is able to achieve 10% of network's maximum throughput, and iSCSI uses 50-60% of bandwidth. 5-20Tb datastores are common nowadays, thus iSCSI could spare you weeks of recovery time. The drawback is that you'll need additional set up before using iSCSI on a standard ESX server. Basic idea is to create Linux guest machine, allow it to use ESX's datastore by mapping RDM disk and start iSCSI target on this Linux VM. All what's left is to set up iSCSI initiator on Windows recovery machine, adding iSCSI disk to the system and you are ready to recover. More details and instructions are here.
Naturally, such set up is not for cases where datastore is saved on a file server or NAS with built-in iSCSI support. At this case just establish a link to the required datastore using Microsoft iSCSI initiator.
We don't recommend using SSH for VMFS disks, as SSH protocol works about 30 times slower than direct connection for single disk. We recommend to use SSH with Uneraser mode to estimate approximately recovery results, unless there is no other way to access SSH.
Boot ESX server using spare HDD with Windows OS
Depending on your case and system configuration, it's possible that there is sense to connect spare HDD with Windows and boot ESX server from it. At this case all VMFS disks became local and recovery process can be run.
Using Raw Device Manager
More sophisticated variant of previous method. You can use use a VM running on the same ESX server where lost data is. To configure VMFS disk as local to guest OS, please follow this guide.
USB adapters and racks for external HDDs
USB adapters or HDD-racks are also out of our recommendation. We had a lot of negative experience when USB adapter read HDD incorrectly and prevented successful recovery. A lot of them have low build quality that affect recovery much more often than anyone would expect.
Server racks for HDDs are normally out of this complain. Just make sure there is no issue that may affect results.
If you don't have enough ports to connect all disks from array simultaneously, consider using “Create Image" and “Mount Image" commands to place several disk images from RAID on single HDD with larger capacity.
After everything is set up, start the software and wait until all disks and arrays(if present) are detected. You should see a disk list and Raid Wizard will start automatically. I suggest you to close it for now and check the disk list first. It is also recommended to switch to Details view to see more information about present disks.
During launch we try to detect all known arrays that have structure information and if they have no issues they should be recognized and rebuilt automatically. If RAID disk doesn't have structure information on it's disks (some controllers store it withing it's RAM) or array is corrupt, at this case RAID should be reconstructed with RAID Wizard.
Normally such RAID failures like broken controller, crashed NAS' power adapter, server shutdown because overheating or simple movement of array to another computer have a good chance to be detected and reconstructed automatically during launch. So check if disk list have required partitions. They should be available at the “Hard Disk Drives" section.
Same deal for any single disk volumes. You can just open required partition at the “Hard Disk Drives" section and browse for required data.
Please refer to RAID Disks section below to learn more about how to use RAID wizard to reconstruct array.