James Allsopp wrote: > I'd like to hear about the optimisations, but I think I'll wait till I get > the system rebuilt! Basically I had expected you to use either rescue mode of the d-i or a livecd or other to assemble the arrays. You did. But neither array came up completely correct. One came up with one disk degraded. The split brain clone came up on md127 instead of md0. The other one came up on md126. So you should fix those using the discussed instructions. I was thinking you would do that from the same system boot that you had posted that information from. But your recent mail implies that you shut the system down and went away for a while. So now it appears you need to "rescue" the system again by the same method you used to get that information you posted. All of that is fine. Except now we already know the information you posted. And so now we know how those arrays are supposed to go together. But that is okay. You can go through rescue mode and assemble the arrays exactly as you did before. And then --stop the arrays and assemble them correctly. But since we know how they are supposed to be assembled now you could skip the assembly of them in rescue mode or livecd mode or whatever you used and simply assemble the arrays correctly the first time. Basically I think you are going to do: * rescue * assemble arrays * stop arrays * assemble arrays correctly Which is perfectly acceptable. The result will be fine. But now that we know what we need to do you could simply do this: * rescue * assemble arrays correctly But I don't want to distract you with complications like this! :-) And then after you get everything working you should visit the partitioning on that second array. Your partitioning starts at the sector 1. But that won't be aligned. It will cause all writes to be a read-modify-write and performance will suffer. > Device Boot Start End Blocks Id System > /dev/sdd1 1 243201 1953512001 fd Linux raid autodetect > Partition 1 does not start on physical sector boundary. > Device Boot Start End Blocks Id System > /dev/sde1 1 243201 1953512001 fd Linux raid autodetect > Partition 1 does not start on physical sector boundary. Instead of using the entire disk starting at 1 it would be much better if you started at sector 2048 as is the new standard for Advanced Format 4k sector drives. I would expect that to be a large performance lever on your system. But fix that after you have your data up and available. Bob
Attachment:
signature.asc
Description: Digital signature