Unsolved
This post is more than 5 years old
4 Posts
0
25646
July 1st, 2006 23:00
PowerVault 124T LTO-2 Veritas 10.d Excessive soft write errors
PowerVault 124T LTO-2 Veritas 10.d Certance tape drive Adaptec 39160 dual port controller Dell Power edge 2800 perc4/di raid controller 2GB ram Server 2003 Veritas tape drive drivers loaded. Bios versions. Autoloader V31.0 Drive 1826
Block 64K buffer 64K number of buffers 10 High water 0
The SCSI cable sent with controller by Dell is 4 Meters long
The moment the unit was installed we have had soft write errors. Veritas writes with compression about 270GB per tape and generates 3500 soft write errors. We have tried
multiple tape brands Dell, Fuji and Maxell. The behavior is the same on all brands.
If we zip folders containing lots of 1k files the number soft write errors go down by 40%.
If we zip all of the data in folder on the server, turn off compression and write 84GB to a tape the number of soft write errors goes down to 118. Same data unzipped written to the exact same tape with compression enabled generates between 991-1065 soft write errors.
Any Ideas ?
Thanks, Brian
DELL-Bob D
899 Posts
0
July 3rd, 2006 14:00
Interesting observations. I am looking into what the limit is for Soft Write errors for that drive are. Will post when I find the answer.
What scsi controller driver and tape drive driver versions are you using?
BrianLord2
4 Posts
0
July 3rd, 2006 14:00
2/4/2004
6.4.630.100
3/17/2005
1.9.0.0
SteveB65
5 Posts
0
July 11th, 2006 12:00
DELL-Bob D
899 Posts
0
July 19th, 2006 18:00
sorry for the delay on this. I still have not recieved any information. (been on vacation for awhile also :smileywink: )
I can tell you that the soft write errors are viewed as a ratio to the data being written. When the ratio of errors exceeds the limit set forth by the drive (the # I am trying to get for you) then the drive will trigger the cleaning. So if your h/w is not requesting a clean you are within the ratio.
This is not a complete answer and I will continue to get the actual # / MB set forth by the hardware vendors..... it is turning out to be a tougher search than I anticipated.
arthuryee
8 Posts
0
July 24th, 2006 05:00
DELL-Bob D
899 Posts
0
August 30th, 2006 21:00
Brian,
Here is the response I recieved about soft errors and LTO...
The LTO format records data in what is referred to as Codeword Pairs and Codeword Quads. A Codeword Quad includes two Codeword pairs and is the smallest unit that can be rewritten. A Codeword Quad holds 936 bytes of data. The ECC system records 64 Codeword Quads were 54 contains user data and 10 contains ECC data. On average a Codeword Quad therefore contain 790 (936 * 54/64) bytes of user data.
Whenever an error is detected in a Codeword the Codeword Quad containing the Codeword in error will be rewritten. The LTO formats require all tracks written in parallel to be synchronized. For this reason the Codeword Quads located on the other physical tracks has to be rewritten as well. For these reasons a rewrite will result in capacity loss of 6320 bytes (790 bytes * 8 tracks) on LTO2 and 12640 bytes (790 * 16 tracks) on LTO3.
Practically the LTO formats have reserved approximately 5 % (10GB on LTO2) extra capacity for rewrites. You will use this extra capacity if there are approximately 7 rewrites / MB on LTO2.
BrianLord2
4 Posts
0
August 31st, 2006 00:00
DELL-Bob D
899 Posts
0
August 31st, 2006 13:00
You are "allowed" 7 errors per MB before you exceed the 5% extra capacity the media vendors have allotted for soft errors.
100GB = 102400MB
102400MB * 7 = 716,800 errors
Although 22500 errors seems excessive, you have only used roughly 135MB of the 10GB allotted for that tape (22500errors * 6320bytes) If you are satisfied with your throughput #'s for the backups I see no reason to worry about these errors at this time.
BrianLord2
4 Posts
0
August 31st, 2006 14:00
The backup is so slow. 17-23 hours when the back up should take 5-6 hours.