I've got a 3TB drive that is failing. It was formatted with GPT table with a 500GB Mac HFS partition and a 2.5TB exFAT partition. I bought another 3TB drive to replace it from a different manufacturer. I ran ddrescue for several days and it was able to read/clone everything except for a few Kb around the 1565-1566GB mark and when I finially killed it, it was down to 29 errors spread through that area.
After ddrescue finished I thought I should be able to just see the cloned disk now since it cloned the partition map too, but Windows 7 says I need to format it before I use it and Ubuntu doesn't list it at all in the file browser (I'm using Ubuntu Live to recover) I thought maybe the partition map had become corrupted so I launched testdisk and did the following:
- Quick search using the default settings. When it started it gave me some chk_fat warnings about heads or cylinders or something not matching. The quick search was taking forever so I let it run over night. When I came back this morning, It had found a small EFI/GPT partition, a ~500 GB Mac HFS partition, a ~250 GB MS partition, a couple really small MS partitions, and then a gigantic MS partition that it told me it could not recover because my drive was smaller than 1000 something TB. I quit that analyze session without writing or doing a deep search.
- Then I noticed that testdisk was using a default sector size of 512 but the original disk had a sector size of 4096. So I changed testdisk's sector size to 4096 and then as soon as I selected analyze the 3 partitions I was expecting (small EFI/GPT partition, 500GB Mac HFS partition, 2.5TB MS partition) all showed up along with their original names. I ran a quick search that finished immediately finding only a 3TB NTFS partion that was on the disk when I bought it. I was thinking it would give me the option to write partition map containing the 3 partitions instead of the partition map with just the single NTFS partition.
- I then started a Deep search hoping it would give me the option to write the 3 partition partition map instead. I'll see what progress it has made when I get home tonight.
As a side question, is it possible for ddrescue to have changed the disk geometry? I can't say for certain, but I thought that before I ran ddrescue, the new disk had a slightly different geometry than the old one, but now the geometries are much more similar. The only differences I can find are the number of cylinders (old is 40,000 something, the new is 37,000 or 38,000 something) and the logical sector size. When using ddrescue, I did use the --sector-size=4096 and --force options.
Sorry if this is verbose. Like I said I new at this and I'm not sure what information would be helpful.
Thanks for the help!