Shulou(Shulou.com)06/02 Report--

[User Unit]

A branch of China Unicom

[Data Recovery Failure Description]

SUN optical fiber storage system, the central storage is RAID6 composed of 6 300G hard disks, divided into several LUNs, MAP to different service servers, the server runs SUN SOLARIS operating system.

Under normal working conditions, users need to add new applications, so an IBM server is added, and then a LUN in the storage is mapped to the new IBM server online. Unexpectedly, the mapped volume is mapped to a LUN on the SOLARIS production system. Since it is not discovered in time, the IBM server has partially initialized this LUN (operation unknown). After that, the disk on SOLARIS reports an error. After restarting, problems are found, and the volume cannot be mounted.

SUN engineers after detection, the implementation of fsck, after the completion of the file system can be hung, but a lot of data loss or size becomes 0, especially the latest data corruption is serious.

Data recovery failure analysis

Such failures are common in SAN environments, but most are caused by human carelessness, as is the case with this failure. Under normal circumstances, LUNs allocated by SAN are in exclusive mode. If they are controlled by several operating systems at the same time, it is easy to cause write operations not mutually exclusive, resulting in file system consistency errors.

If you want to recover this part of the data, you need to go deep into the file system and examine the damage to its various structures. In this example, because the file system uses UFS, for any file that needs to be restored, priority should be given to whether the directory information, nodes, and data areas are normal. If the above three structures are normal, the data can be completely restored. However, in most cases, INODE will be cleared after fsck, even if the directory information is left, it cannot correspond to the data one by one. At this time, it can only refer to the internal format of the file for type recovery.

Data Recovery Process

1. Complete backup failed volume. Because RAID has no failure, dd backup of original LUN is directly done in SOLARIS environment.

2. Analyze the file system in the backup and determine that all inode files to be restored have been removed and cannot be restored. It has to be processed by file type.

3. Analyzing the specific files that users need to recover, we find that the index files using vfs document system have strong type characteristics, and the files contain directory information.

4. According to the index structure characteristics of the official document system, write a program to extract and rename according to the characteristics after extraction.

5. Recover the data file by type, and then the user manually rearranges the data file according to the index file.

[Data Recovery Conclusion]

After 24 hours, 99% of the directory index files were successfully restored, most of the data files were successfully restored, and the rest were destroyed and could not be restored. Users collected from other departments again according to the directory index files.

In conclusion, the user acknowledges that the data recovery was successful.

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.