StoreEver Tape Storage
cancel
Showing results for 
Search instead for 
Did you mean: 

LTO3 drive halting at intervals

berk basarir
Occasional Contributor

LTO3 drive halting at intervals

Hi,

I'm performing some tests with an LTO3 drive controlled via a tape library and shared by a Unix and Windows storage node in a SAN environment. The backup software used is Legato Networker 7.2.

While performing parallel saves using hardware compression I could reach speeds of 77MB/sec.

However the iostat output reveals that the tape drive stopped all activity (meaning 0MB/sec output) for 1 minute around 4 times during a backup that lasted 1 hour. For 1 minute, the drive stops all activity, iostat showing 0MB/sec output. Afterwards the tape continues performing backup at a rate of 75-77MB/sec again.

Same behaviour is experienced both on Unix and Windows, so we are pretty sure this anomaly may be related to tape hardware.The data we are backing up is oracle data, and we use hardware compression on the drive. Can anybody share his experience for this problem?

Thanks in advance,

Bërk
8 REPLIES
Arend Lensen
Trusted Contributor

Re: LTO3 drive halting at intervals

Berk,

Please use the LTT tool to run host and tape drive performance tests.

www.hp.com/support/tapetools

Kind regards,
Arend
berk basarir
Occasional Contributor

Re: LTO3 drive halting at intervals

Hi ,

Thanks for the response.

Actually we have received following results from tapetools report for the save I have described in my previous email:

Datasets Written: 52915
Rewritten Datasets: 4
Rewritten CCQs: 2168267

Now our reasoning is this (based on the reasoning of HP for the HP Ultrium-3 drive):

"#Rewritten CCQs" / ("#Datasets Written" * "#CCQs in a Dataset") is an indication of the quality of the write process. This means
2168267/ (52915 * 128), i.e. 0.32 or 32% which is very bad and seems way to high.

When looking at the global figure of all CCQ retries compared to written CCQs the figure is 0,24 or 24% which is still very bad and proves that the problem is not limited to just this single test.

Further the number of "Rewritten Datasets" corresponds exactly to the number of 1 minute long hangs of the save process during the backup .

Do you think it will still prove useful to test the drive?

Regards ,

Berk
Leif Halvarsson_2
Honored Contributor

Re: LTO3 drive halting at intervals

Hi,
Are you sure the problem is related to the drive ? It could be the backup software as well. Is there more then one drive in the library and, is the behaviour typical for all drives ?
Vincent Fleming
Honored Contributor

Re: LTO3 drive halting at intervals

Always be very careful when you have more than one system zoned to see a drive.

If a backup is in progress on one system, and another queries the drive (ioscan and such), it can cause trouble. I'm not sure about the LTO-3, but I've seen other drives ignore an Identify (what ioscan asks for) because it's busy. Normal behavior for ioscan would be to send a reset to the device and try again if the message was ignored.

A reset would cause the tape drive to rewind. Not good for backups. :-(

Of course, you could also just have a bad tape. Have you tried another?

Regards,

Vince
No matter where you go, there you are.
Curtis Ballard
Honored Contributor

Re: LTO3 drive halting at intervals

You state "LTO3 drive controlled via a tape library and shared by a Unix and Windows storage node in a SAN environment".

This sounds like a bridging environment with the drive hosting a LUN for the robot. Is that the case?

In that environment I have run into several drivers that are unable to have commands outstanding to two LUN's under a single ID so the driver will quit sending data to the drive when there is a command being sent to the robot.
berk basarir
Occasional Contributor

Re: LTO3 drive halting at intervals

Hi all,

Thanks for the feedback. I finally downloaded hptapeperf to test some backups directly writing onto the tape drive from my Solaris server.

I used the following command to create data:
./hptapeperf -i 2 -o /dev/rmt/5cbn -b 65536 -c 16 -r 40960

So used a blocksize of 64KB , 16 blocks at a time for a total of 40GB.

As a result
40960.00 Mbytes transferred in 413 seconds, with average tape speed of 99.18 Mbytes/sec.

However, after 300 seconds there was again a one minute stop, (ie iostat did not report any transfer).

This points me towards a firmware, hardware issue on the LTO drive or an SAN component.

Regards,

Bërk
David Ruska
Honored Contributor

Re: LTO3 drive halting at intervals

> Same behaviour is experienced both on Unix and Windows, so we are pretty sure this anomaly may be related to tape hardware.

The "SAN" is also a possible common factor.

There are a few polling issues from Windows hosts that can cause problems.


Disable Win2000 RSM (also applies to Win2003):
http://h20000.www2.hp.com/bizsupport/TechSupport/Document.jsp?objectID=c00023443

You'll need to check that your legato configuration does not use RSM (most ISV apps communicate via the NT miniport or storport drivers).

Disable Win2003 Plug & Play polling:
http://h20000.www2.hp.com/bizsupport/TechSupport/Document.jsp?objectID=PSD_OT040301_CW01

Can you temporarily zone out all but one backup host, and see if the problem persists? That would rule out SAN interactions.
The journey IS the reward.
Arend Lensen
Trusted Contributor

Re: LTO3 drive halting at intervals

Berk,

Please post a NSR Report here so we can have a look. I can check the NSR Report for polling. Collect it at a moment where there's no backups activity please.
Arend