Shulou(Shulou.com)05/31 Report--

In this issue, the editor will bring you what is the solution for HP EVA4400/6400/8400/P6000 data recovery. The article is rich in content and analyzes and describes it from a professional point of view. I hope you can get something after reading this article.

[antecedents]

HP EVA4400/6400/8400 has entered a high fault zone because of its near retirement. Users tend to think that the eva series purchased at hundreds of thousands or even millions of dollars should be very stable, but in fact, no matter how expensive the storage device is, the disks used are the same or similar. Just like the 1T FATA disk on eva, which sells for part number 454414-001, it is actually the FC version of the infamous Seagate 7200.11 series. Although the former still sells for nearly 4K, while the latter sells for less than 400k, the performance and stability of the two are almost the same. These 1T FATA disks are always damaged in large quantities after 2-3 years of operation, and it is almost difficult to repair once they are damaged. One of the most exaggerated cases I have encountered, the 108T FATA EVA6400, has damaged more than 50 pieces in half a year, which is surprising but normal (due to the design defects of Seagate).

Even FC disks using 15000rpm or 10000rpm are actually equivalent to SAS disks from Seagate, Hitachi and other manufacturers (the current EVA series disks are all based on SAS, and even FC is converted from SAS to FC through conversion circuits, such as 671148-001). EVA is a virtualized storage device, usually data will be constantly migrated, coupled with the application is usually more onerous, so the load of the disk is relatively heavy, but also very prone to failure. EVA relies on the redundant space of a large number of disks and the dynamic migration of rss redundant disks after a failure to achieve data protection of the entire storage, but as more and more disks go offline, this protection will be close to the critical point until it crashes.

The execution of the forced activation command through the fieldservice under eva command view will lead to cache dirty, data block structure inconsistency and data corruption. Second, the repair can only be built on the basis of bypassing the bad road area, if the bad road can not be avoided, it is also futile.

There are other situations: the host can not be connected to the expansion cabinet resulting in data loss, mistakenly deleted vdisk, etc., but also need to have a practical solution.

The HP P6000 series controller is exactly the same as the EVA series internal algorithm and is no longer mentioned separately.

[summary of the principle of eva series storage structure]

EVA series storage is a high-end storage device in HP for the purpose of virtualized storage. The internal structure is completely different from the ordinary storage based on simple RAID. EVA is called VRAID. Including some senior HP engineers, there is actually a misunderstanding of EVA (or it may be the official macro description of HP).

EVA signs each physical disk (PV for short) (written in sector 0 of each disk) and assigns it to a different DISK GROUP. In disk group, similar to aix's volume management, or HP LVM's volume management, each PV is divided into several storage units according to a certain size (according to AIX, let's call it PP). The size of PP is an integer power of 2 and should be between 2 and 16m.

There is a limited number of PP in each PV, and these PP add up to form the free space of the entire DISK GROUP. All PV are composed of several groups of RSS according to 5-15 (the minimum number of RSS disks in HP's official data is 6, the maximum should be less than 15, but the underlying analysis shows that there are 5 and 15 PV RSS cases), each RSS is a so-called regular RAID redundancy group, but this redundancy group is not equal to conventional RAID. Similar to conventional RAID, conventional RAID is a disk-based RAID algorithm, while RSS is a PP-based RAID algorithm.

To improve performance, HP EVA tends to take turns allocating different RSS groups, but the data storage between these RSS is based on JBOD (HP official and many materials are called RAID0, or macro concept), and the members of each RSS are actually PP in different locations in different PV.

No matter how many members there are in RSS, for VRAID5, there are always five PV in a stripe. For VRAID6, there are always 6 PV in a stripe. For example, in the case of VRAID5,EVA, it is possible to combine C (NMagne 5) on N disks in order to achieve IO load balancing.

When a PV in a RSS is offline, the controller looks for available PP from other disks in the same RSS group (in addition to the PV that already exists in the same band), and logically implements the rebuild of each stripe, thus ensuring the security of the entire storage.

When the number of damaged disks in a RSS is small enough (less than or equal to 6), the security of the RSS is very low, and the EVA merges the RSS into another RSS so that the available redundant space is shared and the space can be migrated from another more secure RSS.

In order to ensure that there is enough space to provide redundancy protection, when creating a disk group, EVA provides a protection level of Protection Level. Single means to use 2 disks for redundancy and double means 4 disks for redundancy. But this redundancy is different from hotspare, this redundant space will only be reserved to the tail of each PV, on the one hand, multiple disks, IO will be more uniform, faster; second, the tail space is usually poor quality space.

[common failures of EVA series]

1. Multiple disks in RSS are offline, exceeding the redundancy protection level.

2. When a new disk is added and the data is migrated, there is a physical failure on the new disk (it cannot be rolled back or moved forward at this time).

3. Delete VDISK or EVA initialize.

4. The sudden host cannot connect to the storage. Unable to discover to storage.

[data recovery solution principle]

The core structure of the eva series comes from all vdisk's operational pp tables, which are migrated due to the constant migration of disks. All failures can be recovered based on this map.

When pp map does not exist, according to the redundancy relationship between different stripes, there can be an optimization algorithm to set all PP stripes, thus forming several groups of correct stripe data, and then reassembling several stripes based on file system structure, data structure and other characteristics.

[data recovery resolution process]

1. The original disk image. Unplug the cable at one end of the EVA host and connect it directly to the host HBA card, then you can recognize all the physical hard disks, and then mirror the disk by professional means (winhex under dd,windows under Linux, etc.). As the connection between the eva host and the expansion cabinet is mostly copper wire, it may be necessary to add an optical fiber transceiver module on the expansion cabinet, and then access the fc hba card through the optical link. Of course, all the hard drives can be removed and mirrored in other fibre Channel cabinets.

The connection figure mirrored using the EVA expansion cabinet is as follows:

2. Reassemble the vdisk through the frombyte recovery for hp eva program and write it directly to the mirror file or the target physical disk.

3. Interpret the image file or target disk in the usual way, migrate the image or export the internal file.

This is what the solution for HP EVA4400/6400/8400/P6000 data recovery shared by Xiaobian is. If you happen to have similar doubts, please refer to the above analysis to understand. If you want to know more about it, you are welcome to follow the industry information channel.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.