Ddrescue V Slow (1.10)

[djmuk3]

I am recovering an 80-GB HDD and V1.10 seems much slower than the original
version I was using (1.2?) - currently has been running for nearly 24
hours...

The logic of the old version was fine as it only read a failing sector twice
- once when it tried to read the 'large' block of sectors, then again when
it was reading the smallest possible group of sectors (native block size).

Now I suspect that failed sectors are being read multiple times - unless
there is some cleverness going on with identifying how many sectors were
read successfully on a failed 'large' read... I suspect the 'trim' works by
reading until an error is hit & it looks as though the split process doesn't
always drop to the raw block size but tries to read larger blocks initially?
It is 'expensive' in time to read a failed sector - on the current disk it
is causing IDE bus resets every time a failed sector is hit which is causing
it to run at a snails pace!

Regards

David Mitchell



_______________________________________________
Bug-ddrescue mailing list
Bug-ddrescue@...
http://lists.gnu.org/mailman/listinfo/bug-ddrescue

[Antonio Diaz Diaz-2]

ddrescue.admin@... wrote:
> I am recovering an 80-GB HDD and V1.10 seems much slower than the original
> version I was using (1.2?) - currently has been running for nearly 24
> hours...

Did you try 1.2 and 1.10 on the same HDD?


> The logic of the old version was fine as it only read a failing sector twice
> - once when it tried to read the 'large' block of sectors, then again when
> it was reading the smallest possible group of sectors (native block size).

Ddrescue has never tried to read a sector more than two times. In
version 1.7 a 1 sector read was inserted before every large read to
minimize kernel readahead of bad sectors. This should make ddrescue
faster on bad areas, not slower.


> Now I suspect that failed sectors are being read multiple times - unless
> there is some cleverness going on with identifying how many sectors were
> read successfully on a failed 'large' read...

Of course ddrescue knows how many sectors were read successfully on a
failed large read, and tries to "trim" the rest of the sectors in the
large read only if this is the last bad large read in the current bad area.


Regards,
Antonio.


_______________________________________________
Bug-ddrescue mailing list
Bug-ddrescue@...
http://lists.gnu.org/mailman/listinfo/bug-ddrescue