Data Integrity Validation and Splitting Mirrors

Data Integrity Validation and Splitting Mirrors 08/18/2005 1:32 PM

Page 1 of 3

Data Integrity Validation and Splitting Mirrors:

Best Practice

(Version 3.0, Revised 2005-08-18)

Overview

Splitting mirrors is the practice of bringing down one-half of the mirrored pair of disk

drives. This operation may be performed for a variety of reasons, including disk upgrade,

hardware repair, and data protection. The volume is protected from invalid data due to

checksum and media failures only when both drives in the mirrored pair are available.

One or more distinct blocks of data may be invalid on each of the drives, but when both

drives are fully available, the I/O subsystem can automatically re-route read requests to

the other drive of the mirror in order to overcome such failures.

Good practice should require that all files and tables be validated prior to splitting mirrors,

to detect and correct latent failures.

This document describes the best practice for data validation prior to splitting mirrors. It

covers both G-series starting with G06.24 and all H-series. While the validation process

is meant to reduce the risk of splitting mirrored drives, there is no method available to

eliminate this risk. Splitting mirrored drives creates a single point of failure.

Although disk hardware failures are relatively rare, they do occur, and mirrored

drives should be split only when absolutely necessary. Routine data backup procedures

should be followed to avoid single point of failure data loss.

Prior to bringing down a drive in a mirrored pair, the NSK FCHECK utility should be

used to validate that the data is the same on both the primary and mirrored drives. The

FCHECK –SCAN option will detect checksum errors and correct them when possible,

and validate that the data is consistent across the mirrored pair.

When any read detects a checksum error on either drive, the DP2 driver will

automatically attempt to read the data from the other drive of the mirrored pair. If the

data is valid, the driver will automatically rewrite the data to the drive with the invalid

checksum.

When the FCHECK –SCAN utility is executed, all target data is read on both drives and

compared. If the data on both drives incurs an invalid checksum, the data cannot be

repaired. EMS events will be generated for this case reporting the disk sector and

FCHECK will report the file relative location of the checksum error.

PAGE 1
Data Integrity Validation and Splitting Mirrors 08/18/2005 1:32 PM Page 1 of 3 Data Integrity Validation and Splitting Mirrors: Best Practice (Version 3.0, Revised 2005-08-18) Overview Splitting mirrors is the practice of bringing down one-half of the mirrored pair of disk drives. This operation may be performed for a variety of reasons, including disk upgrade, hardware repair, and data protection.
PAGE 2
Data Integrity Validation and Splitting Mirrors Page 2 of 3 If the volume is mirrored, and both drives of the mirrored pair incur an irrecoverable checksum error at the same location, this may be a symptom of various failures: 1. Path problems – in this case, it is typical for there to be multiple reports of unrecoverable checksum errors in a variety of locations across a variety of files and tables. 2.
PAGE 3
Data Integrity Validation and Splitting Mirrors Page 3 of 3 Recovery from Data Integrity Loss Any inconsistencies reported by FCHECK -SCAN require special handling to determine the correct course of action to properly recover the data set. As a general rule, an FCHECK without the –SCAN option should be performed if the reported file or table is structured. The first step is to identify if one of the disks in the mirrored pair is defective or contains all the inconsistencies.