ADIC tape robot

[Michael O'brien_1]

When I do a search for "dlt adic" on the forums search engine. One of the result returned is "Configure adic dlt on hp" but each time I try to access this response. I get a 404 error document not found.

Does anyone have any information regarding seetings and configuration of an Scalar 1000 robot tape drive on HPUX 11.00 or does anyone have a copy of the "Configure adic dlt on hp" post.

Thanks
Michael

[James R. Ferguson]

Hi Michael:

Due to some changes at the ITRC over the weekend, links were broken. If you get a '404' error when you attempt to access a link from a search, *and* the link contains the string '/1,1150', change it from '/1,1150' to '/1,11866'. I think you will then find what you should.

Regards!

...JRF...

[linuxfan]

Hi Michael,

what Jim is saying is if the URL said

http://forums.itrc.hp.com/cm/QuestionAnswer/1,1150,0x40af6af52b04d5118fef0090279cd0f9,00.html

change it to,
http://forums.itrc.hp.com/cm/QuestionAnswer/1,11866,0x40af6af52b04d5118fef0090279cd0f9,00.html

(See the change 1150 to 11866 in the link)

-HTH
Ramesh

[John Payne_2]

We have an ADIC 1000 plugged into a K380. What exactly do you want to know? BTW, I would not recommend using a K380 to plug into... Are you looking for how to see the drives? Or are you talking about past that point? What backup software are you using?

[Michael O'brien_1]

Thanks, John And Ramesh I manged to get to the posting about the ADIC robot. Unfortuantely it the post was regarding a differnet issue.

John,

We are using Tivoli as our backup software. We have been having issues with tape changing but I think this is a Tivoli issue.

But recently we issue of the speed at which the backups are taking was raised. ADIC and IBM are point the finger at the HP box as being the cause of this issue.

The robot is connected to an L-Class with 1GB of RAM. This box is also being used as an Oracle Express server and as the Tivoli manager.

I have notice that this box is swopping out about 500mb

Do you now if there are any optime setting for Buffer cache for a backup server.

We ran a backup yesterday as a test. It took 2 hours 40 minutes to backup 60GB.

Any pointers or advise would be greater appriciated and help defend the HP corner.

Thanks
Michael

[Frank Slootweg]

As to tape changing, you may want to test the robot and its associated device file, SCSI pass-through driver, etc. with the mc(1M) media changer utility. If it works with mc(1), the backup software should be able to do the same.

As to speed, what type of drive is this?
If DLT7000 then I guess 60GB in 2h40m is not too bad, because (if I calculate it correctly) that is 6.4 MByte/second, while the Native Mode (uncompressed) maximum speed of DLT7000 is 5 MByte/second.

If this would be OmniBack, I would advise to test a local /dev/null file device to rule out network and device bottle necks. Next remote /dev/null file device, next local tape device, etc.. I assume that other backup software has similar functionality.

[John Payne_2]

Michael,

Is this a L1000? L2000? L3000? How do you have the drives configured? Do you have them daisy chained into 1 scsi interface? We have 3 DLT7000 currently daisy chained into 1 SCSI interface on our K-class, and we have found bandwidth issues on the box. Single stringing the drives is optimal, but we think we will be able to get away with 2 to a scsi inteface when we move to an L-class at the beginning of the year...

If you are swapping that heavily during your backups, (or at any other time) you have big problems... I am sure that thrashing is problibly interfering, since your system is already most likely quite busy with Oracle Express. Also, your backups will try to write their indexes, and you may have disk bottleneck with all the paging that is happening.

What are the values of dbc_min_pct and dbc_max_pct? If you drop dbc_min_pct to an approppriately low value, it should shrink as needed. You should also be able to tell approx. how much you really use, and then set the values appropriately to free some memory. Here is a thread that talks extensively about buffer cache and how to tune it. http://forums.itrc.hp.com/cm/QuestionAnswer/1,11866,0x997cabe92dabd5118ff10090279cd0f9,00.html Was the test backup you ran only writing to 1 tape, or was it writing to multiple tapes? If it was 1 tape and it is DLT7000, then you have a good write rate for that test backup.

Has ADIC upgraded the firmware on all the drives? We upgraded to 101 level a few months back, and this seemed to clear out some issues with errors out of the tapes. Also, we used to have two different types of DLT7000 in our Emass unit, and we had them homogeonize our environment with the same drives.

John

[Michael O'brien_1]

Hi John,

I think I missed off some important information, The machine is an L2000, the tape drive in the ADIC ronot is an IBM Magstar 3950.

The dbc_min_pct = 5 and dbc_max_pct = 10

We are having an extra 1 Gb installed at the weekend. I just found out that we have an identical robot attached to an N-Class which currnetly is being used for DR. So I'm going to run some tests on this box

John thanks for your feed back, it's much appriciated