About Verification

For the 3ware 9000 series, the verify feature combines verification of redundant units (confirming the validity of the redundant data) and media scans for non-redundant units. (3ware software for the 7000 and 8000 series provided separate features for verify and media scan.)

Verification is useful because it can provide early warning of a disk drive problem or failure. For example, if the verification process encounters many error locations, it may be an indication of an excessive number of grown defects on the drive. This allows you to replace drives before they fail.

For a RAID 1 or RAID 10 unit, a verify compares the data of one mirror with the other. For RAID 5 and RAID 50, a verify calculates RAID 5 parity and compares it to what is written on the disk drive.

Verification checks each sector on a drive. This is important, because day-to-day use of the media may leave many sectors on a drive unused or unchecked for long periods of time. This can result in errors occurring during user operation. Periodic verification of the media allows the disk drive firmware to take corrective actions on problem areas on the disk, minimizing the occurrence of uncorrectable read and write errors.

About Verifies of Non-Redundant Units

Verification of non-redundant units (single disks, JBODs, spares, and RAID 0 units) read each sector of a drive, sequentially. If a sector can't be read, it is flagged as unreadable, and the next time the controller writes to that location, the drive reallocates the data to a different sector.

About Verifies of Redundant Units

Verifies of redundant units also read each sector, working from lowest block to highest block. If verification cannot read data in a sector, dynamic sector repair is used to recover the lost data from the redundant drive or drives; this recovered data is written to the problem sector. This forces the drive to reallocate the defective sector with a good spare sector.

How Errors Are Handled

Verification makes use of the same error checking and error repair techniques used during ordinary use of drives configured through 3ware RAID controllers.

When verification encounters an error, the controller typically retries the command. If there are cable CRC errors, there may be multiple retries including downgrade of the UDMA mode. If the error persists and is unrepairable (e.g., ECC errors), an error notification is issued to indicate the problem. (See AEN 026h Drive ECC error reported.)

If the disk drive is part of a redundant array that is in a redundant state (not DEGRADED or REBUILDING), then Dynamic Sector Repair automatically rewrites the redundant data to the error location to force the drive to reallocate the error location. A notification of repair is posted. The result is a restoration of drive and data integrity; the primary and redundant data are again both valid.

If the array is not redundant, a file-system check is recommended to correct the issue. If the errors persist and cannot be overwritten from a backup copy, perform a final incremental backup. You will need to replace the defective drive, recreate the array, and reinstall the data.

Verifying Units

Copyright (c) 2004-2006, AMCC