Tape Alerts and HP LTO drives ?

[Marcel Guwang]

Hello,

I noticed on many HP ultrium drives that they newer issue a READ_FAILURE (05) or WRITE_FAILURE (06) tape alert even when they report uncorrectable errors in the log pages.

Most of the TA are optionnal, but the 5 and 6 are mandatory and should happen on uncorrectables. Any thoughts ?

IBM drives are much more verbose on these Tape Alerts and on some others optionnal Tape Alerts.

Thanks

Marcel

[Curtis Ballard]

The tape drive specifications show those flags as supported on LTO-3 and LTO-4 (I didn't check LTO-2).

The specs indicate that those flags are set when any one of the following ASC/ASC combinations is reported to the host:
0x1400 - RECORDED ENTITY NOT FOUND
0x1100 - UNRECOVERED READ ERROR
0x0C00 - WRITE ERROR

The specs also indicate that the flag will be cleared on a successful unload.

Are you reading the flags before the unload?
Are you receiving one of those ASC/ASCQ values reported in a CHECK CONDITION?

[Marcel Guwang]

Thanks Curtis,

My problem is that I want to monitor the drives by polling the tape alert log page.

I cannot read the ASC/ASCQ buffer, as this one is intended for the backup software.

And it appears that the only tape alerts I could get from an HP ultrium are write protect (09h), No removal (0Ah) and a purposely generated Cooling Fan Failure (1Ah)

regards

Marcel

[Curtis Ballard]

What drive type are you using?

TapeAlerts are also intended for the application and with LTO-1, LTO-2, and LTO-3 drives you can't poll the TapeAlert log page or the application will never get the TapeAlert messages. That is the way the standards designed that page.

It's possible that you are polling the page after the application has already read the TA log page in which case it would be cleared already and you shouldn't see those flags.

If you are using an HP LTO-4 drive then there are some new TA functions in the SSC-3 draft standard that the drive has already implemented. With LTO-4 drives you can read the "TapeAlert Response Log page", 12h, which is not cleared on read and can be read by multiple applications. The flags in log page 12h will only be cleared when the event that caused the flag is cleared. Many of those flags are cleared on an unload. See the SSC-3 standards document at www.t10.org for more information.

[Marcel Guwang]

Curtis,

Through our customers we are seeing all types of drives ranging from LTO1 to 3 (even 4) from any manufacturer.

our Storsentry product is monitoring big datacenters with tenths or hundreds of drives.
http://www.hi-stor.com/site/index.php?pge=62〈=en


When Storsentry is running, the tape alert recovery from the backup application is intendly turned off.

My issue is that part of our drive/media quality measurement is based on Tape Alerts. And statistically, I get nearly no TA from HP drives, while I get much more from other brands. Which makes quality ranking more difficult.

I could agree that they are running better, but even in case of uncorrectable counts in the log pages no TA is issued. In my Lab on a HP-LTO3 I wasn't able either to get one TA 5 or 6 despite severe misabuse! I was nevertheless able to generate a 1Ah TA (Cooling Fan Failure ) prooving the functinnality of TA retrieve.

Regards

Marcel

[Curtis Ballard]

You can post an L&TT support ticket from your test drive and I could check and see if those TA flags should have fired. Just having an uncorrectable error in the logs doesn't mean you'll get a TA 5 or 6 since the drive could reposition and retry and be successful on a follow on read so the host read/write operation doesn't fail.

I've been running lots of HP LTO drives since the first proto of LTO-1 and haven't seen an event that would trip flags 5 or 6 in several years so I can't say that I know the flags work.

[Marcel Guwang]

Thanks Curtis.

If you are interested, I can try harder on my lab drive (erasing specific tracks) and send you the LTT log.

You didn't see some for years, and that's my point, they don't happen very often. But don't worry, I have seen a very few, so they do work (on LTO2 at least).
I just wanted to make sure that I wasn't missing something, seems not as your's are rare too.

Marcel

[Curtis Ballard]

To be clear, my testing would capture a hard read or write errors returned to the host as well as TA flags and I haven't had any hard errors in several years. I mostly work with libraries and tend to wear out the load mechanism before the heads. Even that takes a lot lately though.