LTO-2 Ultrium tape fullness again !!!

[brackovic]

Hi all, especially those who have already replied on my similar threats,

Well, as the subject says, my customer still does have same problem: within same tape library, with same drives, with same tapes and with same drivers for tape library and tape drives, one NW Group (with clients that backup Oracle servers with RMAN) writes more than 300 GB on one tape and another NW Group (with clients that backup ordinary file systems) writes about 120 GB on one tape.

Here and on other sources I have found several possible causes of this problem, but no one solved this particular situation (it was about drivers, HW or SW compression, etc. etc.).

605, you explained me technically NW’s writing end of tape in the case of error. I know for that issue and it’s very strange that this situation occurs and every tape used for File system group backup. And in the same time, every tape used for Oracle group backup, for example doesn’t have such states.
You also explained me that Oracle is much more compressible than ordinary data in file system. OK, it’s very clear for every IT professional, but the differences are too large: if Oracle group saves 318 GB and File System saves 131 GB it means, as we assume that it we calculate with nominal 200/400 GB capacity that Oracle compresses 1.59 and File system 0.655!!??
What does it mean??

Here is the part of the recent deamon.log where there are three “end of tape” logs: for two “file system group” tapes (named Tjedan.009 and Tjedan.010) and for one “Oracle group” tapes (named Oracle.002).

Obviously NW recognizes tapes as LTO-1 tapes, because it stands: “…used 116 GB of 95 GB capacity
…” and “…used 318 GB of 95 GB capacity …”, doesn’t it??

The hardware is Fujitsu MTC Changer LTO 9084-221 with two new LTO-2 drives.
The server where NW Server is installed is IBM server and for all hardware components the newest hardware are installed.

As you can see, File system group recognizes end of tape after 116 (or 131, or similar) GB of 95 GB and Oracle group recognize end of tape for same tapes, same drives, same library and same NW environment after 318 GB pf 95 GB !!!???!!!

Does anybody have some useful idea or suggestion to solve this problem, or to explain logically such kind of behavior???
Thanks,
S

Part of recent deamon.log file:
------------------------------------
03/23/05 02:31:32 nsrd: media warning: \\.\Tape1 writing: unknown error 1100 (0x44c), at file 138 record 10491
03/23/05 02:31:32 nsrd: media notice: LTO Ultrium tape Tjedan.009 on \\.\Tape1 is full
03/23/05 02:31:32 nsrd: media notice: LTO Ultrium tape Tjedan.009 used 131 GB of 95 GB capacity
03/23/05 02:36:08 nsrd: media info: verification of volume "Tjedan.009", volid 3987898 succeeded.
03/23/05 02:36:22 nsrd: write completion notice: Writing to volume Tjedan.009 complete
03/23/05 02:36:22 nsrd: media info: suggest mounting Tjedan.010 on sazbsus for writing to pool 'Tjedan'
03/23/05 02:36:22 nsrd: media waiting event: Waiting for 1 writable volumes to backup pool 'Tjedan' tape(s) on sazbsus
03/23/05 02:36:23 nsrd: \\.\Tape1 Eject operation in progress
03/23/05 02:37:08 nsrd: media info: loading volume Tjedan.010 into \\.\Tape1

03/24/05 01:50:53 nsrd: media warning: \\.\Tape1 writing: unknown error 1100 (0x44c), at file 126 record 9901
03/24/05 01:50:53 nsrd: media notice: LTO Ultrium tape Tjedan.010 on \\.\Tape1 is full
03/24/05 01:50:53 nsrd: media notice: LTO Ultrium tape Tjedan.010 used 116 GB of 95 GB capacity
03/24/05 01:55:34 nsrd: media info: verification of volume "Tjedan.010", volid 4215096970 succeeded.
03/24/05 01:55:46 nsrd: write completion notice: Writing to volume Tjedan.010 complete
03/24/05 01:55:46 nsrd: media notice: Jukebox 'EXABYTE' only has 10 GB available on 1 volume(s)
03/24/05 01:55:46 nsrd: media info: suggest relabeling Tjedan.011 on sazbsus for writing to pool 'Tjedan'
03/24/05 01:55:46 nsrd: media waiting event: Waiting for 1 writable volumes to backup pool 'Tjedan' tape(s) on sazbsus
03/24/05 01:55:48 nsrmmd #4: Start nsrmmd #4, with PID 2708, at HOST sazbsus
03/24/05 01:55:50 nsrd: nsrjb notice: nsrjb -j EXABYTE -O4 -l -R -M Tjedan.011

03/24/05 05:54:16 nsrd: media warning: \\.\Tape0 writing: unknown error 1100 (0x44c), at file 345 record 14027
03/24/05 05:54:16 nsrd: media notice: LTO Ultrium tape Oracle.002 on \\.\Tape0 is full
03/24/05 05:54:16 nsrd: media notice: LTO Ultrium tape Oracle.002 used 318 GB of 95 GB capacity
03/24/05 05:56:56 nsrd: media info: verification of volume "Oracle.002", volid 809221129 succeeded.
03/24/05 05:57:09 nsrd: write completion notice: Writing to volume Oracle.002 complete
03/24/05 05:57:09 nsrd: media notice: Jukebox 'EXABYTE' only has 10 GB available on 1 volume(s)
03/24/05 05:57:09 nsrd: media info: suggest relabeling Oracle.003 on sazbsus for writing to pool 'Oracle'
03/24/05 05:57:09 nsrd: media waiting event: Waiting for 1 writable volumes to backup pool 'Oracle' tape(s) on sazbsus
------------------------------------------

 

[comtec17]

You will get low compression on Filesystems when:

1. The data is to small to compress therefore the compression will blow up (Make it bigger)

2. RMAN/Oracle will compress much easier than Filesystem data due to the large blocks of data.

I have experienced this not only with Legato, but also Veritas, Tivoli, NetVault, and Brightstor.

 

[nokker]

Hi

Have you checked if the media type maybe have been manually set ?

/Tommy

 

[605]

The value of 95GB refers to the "Volume dafult capacity" which is indirectly selected by the device type. Obviously you selected the wrong one.

This is usually not an error as NW writes to the physical end of tape. However, with the device type, the block size is also indirectly selected. This may result in the fact that you use smaller block sizes which results in a smaller capacity. This may be one reason.

If the error code 0x44 is unknown to NetWorker, hat do you find about it in the literature?

 

[RickCald]

Oracle table spaces are very compressable - when I do a cold backup and compress the tables afterward they compress as much as 70%+ so I wouldn't think it would be out of line for the tapes holding oracle data would hold a lot more data than general data on a tape.

 

[databoy26]

This might be a dumb question, but do you have software compression turned on for that specific client? When I had it turned on it caused lots of issues.

btw - I am doing mostly nfs based backups and I probably average about 500 to 600g per 200g lto2 tape. This is mostly oracle data files.