See RAID0+1 and raid1+0 through a data catastrophe

Recently, a case of 4 disk SCSI raid0+1 data recovery, composed of 4 pieces of 36G SCSI. The client said it had done two sets of RAID1. After the failure, 3 disk offline in the RAID state.

According to my understanding, this should be two sets of logical disk (RAID1), that even if it is 3 disk offline, there should be a set of logical disk can work properly. However, the logical disk of the array is not recognized when the customer accesses the array with Windows installed on another hard disk. In this way, many of the explanation is not enough, can only be carefully analyzed.

Take down the hard drive, connect to the SCSI adapter alone, enter the system, no exception, can identify 4 physical hard drives. Analysis, no significant raid information area, after the comparison of 4 disks, the conclusion is that 1, 3rd and 2, 4th, each group has the same sex, but there are a lot of inconsistent data. There are partition tables in plate 1th and 2nd, each of which indicates that the original logical disk partition totals approximately 68G. Accordingly, the following three kinds of conditions are known:

1, two groups of RAID0, but 1, 3rd and 2, 4th are part of the exact same data, should be able to exclude.

2, raid1+0 (that is, 22 do RAID1, and then do RAID0, this security level is high, the customer is the integrator to do, the most likely), for a period of time, the two groups of RAID1 have a hard drive offline (since then the equivalent of RAID0, no longer provide any redundancy). Later, another hard drive went offline and the system crashed. This situation is very much in line with raid performance.

3, raid0+1 (that is, 22 do RAID0, then do RAID1, this is not very good, inference is not likely)

Based on the analysis, it is found that the raid consisting of 1 and 3 is no error and should be correct. Reorganize the data. Write back to raid directly and the system will start normally. File access is also normal.

Originally thought to have been perfectly settled. Results in a very short time received customer calls, said data serious lag, is something two years ago.

A thin thought, dawu.

The real situation should be: the user did the raid0+1, the result is composed of one of the RAID1 in the RAID0 of a disk offline (should be 1 or 3), resulting in the entire RAID0 offline (two offline), and then has been a single RAID0 way to work (remember unexpectedly two years, Khan!) Until recently, one of the remaining groups of RAID0 was offline, and raid was completely paralyzed. The raid card used by the user is a Adaptec 0-channel RAID card, which can not safely buffer the data, and when it is off-line, the file system consistency is problematic because the data part is not written.

Re-organizing 3 and 5th, correcting errors and restoring data 100% to success.

This case highlights the security differences between Raid0+1 and Raid1+0, let's talk about it carefully.

Raid0+1:

Structure, two or more (including two pieces) hard disk to do first strip (RAID0), composed of the same two groups of first-level logical disk. Then mirror the two sets of logical disks (RAID1). The following figure:

Raid0+1 Redundancy (Security): As long as there is a disk error, its RAID0 will be overall off-line, can only rely on the outermost RAID1 redundancy to support. In fact, only one disk can be allowed to go wrong, so that if you have more than 4 pieces of hard disk, security is actually much worse.

Utilization Rate: 1/2