Asset ID:	1-71-1015360.1
Update Date:	2010-12-22
Keywords:

---

Solution Type  Technical Instruction Sure

Solution  1015360.1 :   Lxxx/9740 - How Does Cleaning Work with Autoclean or TapeAlert for Netbackup and DLT7000/8000 Drives  


PreviouslyPublishedAs
223496

Applies to:

Sun StorageTek L1400 Tape Library - Version: Not Applicable and later    [Release: N/A and later]
Sun StorageTek L180 Tape Library - Version: Not Applicable and later    [Release: N/A and later]
Sun StorageTek L20 Tape Library - Version: Not Applicable and later    [Release: N/A and later]
Sun StorageTek L40 Tape Library - Version: Not Applicable and later    [Release: N/A and later]
All Platforms
Reviewed for relevance 22-Dec-2010.

Goal

How Does Cleaning Work with Autoclean or TapeAlert for Netbackup and DLT/SDLT Drives.

Solution

Steps to Follow
There are two types of "cleaning" that can happen on a DLT/SDLT tape drive:

1. A "Cleaning Request", SCSI sense key 01 ASC 80 ASCQ 02, can only occur during the calibration process on tape load.  This is a "Recovered Error" so the tape eventually was able to calibrate and load, however there was some problem during the process that exceeded the thresholds.  If a significant number of these are being generated by a certain tape then this is a indicator that this tape media may be a problem.  If cleaning requests are being generated by multiple tapes then there could be a media management issue, where the media at the site needs to be checked.  This could have been a result of tapes written to by pre-V95 drive code, in which there were known issues.

The wearing of drive heads, problem described above, was corrected with the release of v106 code. Quantum changed the handling of the "Cleaning Request" so that when these sense key 01 ASC 80 ASCQ 02 are sent, they wait until 100 tape motion hours have passed before issuing the request to the library.  However, v106 introduced some additional algorithms for calibration that have caused it to take longer to load a tape and become ready.  This seems to occur on tapes that were previously initialized on drive code that is pre-v95 code. Although it does not seem to affect the operation of more well known backup applications such as NetBackup or Tivoli (longer load time means a longer backup window), it does seem to have an affect on some other, lesser used backup applications.  Some of these applications seem to have specific timeout values on certain operations, so the longer load time causes the backup job to fail. This also means they are not very recovery tolerant.  However, cleaning up any pre-v95 code media would probably eliminate any problems encountered.  Drive code v107 does not have the additional algorithms, but does contain the fix for the Cleaning Requests.

2. The other type of "cleaning" that can occur is a "Cleaning Required". A Cleaning Required is a SCSI sense key 03 ASC 80 ASCQ 01. These are usually seen in a read or write operation and are a true medium error. This is an unrecovered error and is sent out on the SCSI bus to the host so that the backup application can start its error recovery procedure. A Cleaning Request does not do this, it only informs the library that a cleaning may be in order because of out of thresholds on the calibration of the tape. A Cleaning Required is performed immediately upon the drive, where a Cleaning Request is deferred until the drive is free. If another tape is loaded between the request and the drive free, which does not have a problem with calibration, then it will turn the light off.

Quantum's recommendation for cleaning DLT drives is to let the drive decide when it needs to be cleaned. If there are excessive amounts of cleaning requests, then this is usually an indicator that there are media issues that need to be addressed. To eliminate problems with excessive head wear on the drives and allow auto-cleaning on the library, tapes that were initialized the first time with pre-v95 drive code may have to be retired and reinitialize with new drive code in order to rewrite the calibration tracks (probably a necessary step as there were lots of problems that could be associated with pre-95 code). Recommend upgrading the drives to v107 code to implement "Cleaning Request" improvements.

The 84 errors, found in logs from Veritas NetBackup, are the "Cleaning Required" errors. These are passed back to the host on the SCSI bus and are also sent back in the TapeAlert as flag 20. "Cleaning Requests" are not sent to the host and this is why the light comes on, but there is no indicator in Veritas and are not sent back in the TapeAlert. With v106, a clean will not occur until 100 tape motion hours have passed, so you will not see a cleaning light on either.




Auto Clean, Veritas NetBackup, TapeAlert, L20, L40, L80, DLT7000, DLT8000, DLT, SDLT, 9740, L700, L180, How does cleaning work with Autoclean or TapeAlert for Netbackup and DLT drives, Proper usage of library Autoclean or TapeAlert for Netbackup and DLT drives, Cleaning light on DLT7000/8000 drive comes on and no cleaning action is observed
Previously Published As STKKB4474

Attachments

This solution has no attachment