Re: Burning and mounting written BDs

[Thomas Schmitt]

Hi,

Kjetil Kjernsmo sent me the 600+ KB log of a xorriso -check_media run.
I quote from it:

> xorriso : NOTE : Loading ISO image tree from LBA 0
> ...
> xorriso : UPDATE :   40115 nodes read in 9 seconds

This means it should be mountable (after due waiting for the drive to
settle after loading the tray).

Next the whole data range of the loaded filesystem gets scanned
sequentially:

> xorriso : UPDATE :      32 blocks read in 2 seconds , 0.0xB
> xorriso : UPDATE :    1472 blocks read in 3 seconds , 0.7xB
> xorriso : UPDATE :    6080 blocks read in 4 seconds , 2.1xB
> xorriso : UPDATE :   10688 blocks read in 5 seconds , 2.1xB
> ...

Speed increases to 8x BD speed (~35 MB/s) until about 25 GB were read:

> xorriso : UPDATE :  12896k blocks read in 1020 seconds , 8.0xB

Then it slowly decreases (as if reading wanders back from outer rim to
the center):

> xorriso : UPDATE :  14517k blocks read in 1118 seconds , 7.5xB

But at about 30 GB speed begins to wobble and long lasting reads occur:

> xorriso : UPDATE :  14565k blocks read in 1121 seconds , 7.4xB
> xorriso : UPDATE :  14579k blocks read in 1123 seconds , 3.0xB
> xorriso : UPDATE :  14595k blocks read in 1124 seconds , 7.4xB
> xorriso : UPDATE :  14601k blocks read in 1127 seconds , 1.0xB

Note the missing seconds 1122, 1125, 1126 which were spent waiting for
the reply to a single 64 KiB read request.
Eventually errors occur:

> xorriso : UPDATE :  14648k blocks read in 1136 seconds , 1.1xB
> libburn : SORRY : SCSI error on read_10(15000832,32): [4 3E 02] Drive error. 
> Timeout on logical unit.
> xorriso : UPDATE :  14649k blocks read in 1146 seconds , 0.1xB

I am undecided whether it is the drive firmware or the kernel driver
which emits this error code [4 3E 02]. Whatever, the drive has problems
with the medium.

A few more blocks can be read but the error comes back several times:

> libburn : SORRY : SCSI error on read_10(15001696,32): [4 3E 02] Drive error. 
> Timeout on logical unit.
> libburn : SORRY : SCSI error on read_10(15002720,32): [4 3E 02] Drive error. 
> Timeout on logical unit.

Then comes an error which clearly stems from the drive firmware:

> libburn : SORRY : SCSI error on read_10(15004480,32): [3 11 05] Medium error. 
> L-EC uncorrectable error.

After many of these, the run gets aborted due to a safety limit which
i installed in xorriso:

> xorriso : NOTE : -check_media: Reached time_limit=28800
> xorriso : UPDATE :  14859k blocks read in 28801 seconds = 0.2xB

8 hours is more than i expected as human patience from Kjetil.
I hope the drive is still well.

In the beginning of the problems, the unreadable regions were only 64 or
128 KiB of size. But about 120 MB further on, the unreadable regions grow
to about 400 KiB. The unreadable sizes grow to several MB after 200 MB
more of slow and bumpy reading.

-----------------------------------------------------------------------
Summary:

Even if the medium is probably mountable, about 30 percent of the data
file content is in the range that causes errors or is in the yet untested
range at the end of the medium.

You may learn the names of the data files in the problem area by:

  xorriso -indev /dev/sr0 \
          -find / -sort_lba -lba_range 15000832 999999999 \
                  -exec report_lba --

It is possible that other drives can read the problematic blocks.
But unless the currently used drive is clearly defective, one has to
declare this BD-R medium unfit for archival.

-------------------------------------------------------------------------
Proposals for burning a new annual backup without wasting too much money:

- Use single layer 25 GB M-disc BD-R and split your backup into two.
  (That's how my project scdbackup began in 1998 when i had to split
   my periodic backups over more than one CD-RW media.)

- Use non-M-disc 50 GB BD-R and make two or three identical copies.
  First let xorriso create an ISO image file on hard disk:

    xorriso -for_backup -outdev my_big_buffer.iso -map ...

  Then burn the image file onto the first BD-R:

    xorriso -as cdrecord -v -eject dev=/dev/sr0 my_big_buffer.iso

  Then checkread it:

    xorriso -for_backup -indev /dev/sr0 -check_media time_limit=7200 --

  If it turns out good (for the moment) then burn the next copy and
  verify it.
  (Finally remove my_big_buffer.iso from hard disk.)

  Years later you may use xorriso -check_media to check the media again.
  If errors occur, then -check_media runs with options
  "data_to=image_file" and "sector_map=map_file" will try to collect
  copies of the readable media blocks in an image_file on hard disk.
  The more media copies you have, the better is the chance to get all
  blocks read.
  (If the medium is not mountable, use -outdev instead of -indev and
   set option "use=outdev" in the -check_media run.)

I myself would use "25 GB" non-M-disk BD-R media and invest the saved
money into multiple copies. I would probably make one copy on BD-RE
just to have a wider range of chemical media properties.

If my hard disk were too full for an intermediate 25 GB image file, i
would risk subtle differences between the copies and burn multiple backups
with identical input files. If no changes happen with the input, then
a patchwork of blocks from the various BD-Rs should still be a valid ISO
with the correct data file content.
After all, chances are good that a normal BD-R stays readable for decades
if its initial read quality is sufficient.


Have a nice day :)

Thomas