Almost always, when enterprise storage equipment fails, this failure occurs in succession (a disk failure followed by a rebuild, and then a failover that does not complete cleanly). Frequently, the individual disks that make up the failed storage array are functional; however, what has been lost is the configuration (map) that described how these disks relate to one another. Rebuilding the storage array proceeds from the lowest level upward through all layers added by the vendor, and each level relies upon the correctness of the one beneath it. Therefore, having the full shelf in its original order matters more than any single component.
When storage arrays fail, they do so sequentially. Thus, to recover data contained within a storage array, the sequential events that occurred during the time the array was functioning improperly must be unwound in the order they happened.
A typical SAN-related incident is not catastrophic. Instead, most SAN-related incidents start with a disk failure. This is typically followed by a rebuild commencing. Next is usually a secondary disk failure due to the increased workload caused by the rebuild. Then, the system will attempt a controller failover without successfully completing the process. Finally, someone may decide to cycle power to the entire storage shelf in an attempt to resolve any potential issues. Each of these steps can be considered logical and rational. However, collectively, these steps lead to a storage environment where the combination of multiple devices cannot assemble together into usable configurations.
What sets this apart from standard RAID is that there are layers. The physical disks form a RAID set; the RAID set then feeds a storage pool; the storage pool can be partitioned into multiple LUNs, usually thin provisioned; each LUN has either a file system or a datastore. The recovery process must go upward through all the levels in order. Each level cannot move forward until the previous one has been rebuilt properly. Therefore, if you get the pool layout incorrect, then the LUNs above it will have no relevance.
Disk order, stripe pattern and parity configuration across all the member disks. These three are the same as with any other type of array; however, they are frequently spread across several shelves instead of just within one chassis.
How the RAID groups merge into a pool, where the vendor wrote its own metadata, and which reserved areas exist within the pool. This is where having an understanding of the vendor's specific technology is at its highest value.
Which pool extent(s) belong to which LUN. This mapping of each extent in a thin-provisioned data store to the correct logical unit number (LUN) is the most important piece of information, because without it the pool becomes simply an unorganised collection of blocks, with no method of determining which block(s) are part of which logical disk.
NTFS, VMFS, ext4 or a datastore for virtual machines. They cannot be accessed until each of the layers below them has been properly reconstructed, which is why this is the final step that will be taken, not the first.
Almost all of the SAN-related work in this region comes from two different areas. Genuine enterprise storage behind WMS/ERP systems in warehouse management environments exists in Milton Keynes and the northern distribution estates, and when either a controller pair fails or a LUN goes offline, the volumes being affected are in the order of tens of terabytes. The second area is institutional: research computing estates where a storage tier has separated itself from the associated metadata.
Both are reconstruction tasks as opposed to repair tasks. The disks are frequently in good health, but the map to those disks has been lost. Send all the disks from the shelf, each labelled with its slot designation, along with any configuration documentation you still have.
Every member, every shelf. Because enterprise configurations stripe across the entire set of disks, several platforms will not build using fewer than the total number of members. The most common reason SAN jobs stall during the beginning stages is an incomplete set of disks.
The cost of recovering your enterprise storage starts at £1,250 +VAT, and costs above that are then quoted on an individual basis, depending upon the number of members, the size of the storage and the type of platform, which vary far more here than on single arrays.
One thing that does remain consistent is how the billing process works: there will be one written agreement for a price prior to commencing the project, and the price cannot change after the start of the project. There is no percentage of the data and no hourly rate.
Then the failure is in the mapping (data location) rather than the media (disks), which is the usual case. Since we are dealing with mapping failures, thin-provisioning tables and pool metadata are typically the two main things that need to be rebuilt, and this can be done if we have access to all members.
Almost certainly not. Configuration, and not data, is what is held by controllers. The volumes will be reassembled based upon the on-disk layout without their original hardware being required, and that is exactly what this work consists of.
Each disk (yes, each individual disk), labelled with its original shelf and slot. There are several storage platforms that stripe metadata across the entire group and will not allow assembly of the volume unless every member of the group is available.
Once the group has been assembled, there is one additional layer of reconstruction. The sequence runs through the pool, the LUN, VMFS and, finally, the VMDK files contained within it. The virtual machines can then be extracted and repaired as guest systems rather than as simple files. Therefore, they can actually be booted, rather than just existing.
Typically, 10-20 working days represents the time frame for completion. It does take time to image tens of terabytes of data. Reconstruction cannot commence until each member of the group has been successfully imaged. A realistic estimate for completion is provided with the written verdict, rather than an overly optimistic one.
All work is completed in-house. No sub-contracting occurs. Data will never leave the United Kingdom. One named engineer will complete your job from start to finish. Prior to receiving your equipment, we will sign your own confidentiality terms.
We quote £1,250 +VAT as a minimum price per array, and prices will vary depending upon number of members, shelf configuration and vendor format. As such, quotations are provided at scope level rather than list level. A 50% deposit is applicable, with the balance payable upon successful completion.