On 17/05/21 07:28, Roman Mamedov wrote: > As for recovery, you really might need to play with --create; to prevent the > chance of data loss, there's a way to experiment using "overlays", keeping the > original drives untouched; see [2] for more background on that, and [3] > provides steps for your actual situation. Don't forget to use "--assume-clean" > and "--readonly". Firstly, what I'd do is a hexdump of the fifth kb of each disk (ie hopefully where the superblock is/was). Post that here and see if anyone can decode it. It does look like something has created a gpt, so mdadm is no longer looking at the raw disk for the array. There MIGHT be enough information lying around for someone to tell you what --create command to use. Another thing - are those the original disks used to create the array? Have you modified the array in any way? If you haven't modified the array, then a plain create using the same version of mdadm that created the array should just put things back the way they were. The reason you mustn't just use --create willy nilly is that any modification of the array might move the data offset, and different versions of mdadm may have different default settings. The other option is BACK UP YOUR HARD DRIVES, delete the new gpt, and then see if it will assemble. I know I've seen this before, but I've a nasty feeling the previous time the user had used partitions and the GPT had been trashed so it's not quite the same. Still, you might well be lucky ... Cheers, Wol
