NAS-134309 / 25.10 / Make data partition slightly smaller than the disk #15825
Conversation
|
time 8:00 |
bugclerk
changed the title
|
I'll review this in a bit but pyparted actually has a critical issue that I discovered yesterday in 6.12 kernel. It's using the optimal_io_size as the factor for the starting offset of the partitions. That's wrong and so we have a situation where drives that were partitioned with a starting offset of 1MiB, now have different start positions....on these SAS disks, it's now starting at 16.8MiB. I say all that to say that I agree we should remove this program. It's unpredictable and doing wrong things. However, I have been sitting on code for awhile where I used sgdisk source code as a reference to write a python equivalent that doesn't use subprocess. I'll send that to you because I think we should do it here. We're selling increasingly larger systems (many 100's of disks) and we're being pushed to make things efficient. |
|
@yocalebo do you plan non-subprocess disk formatting for 25.04? |
No, probably not. Would rather do that for 25.10 only |
Reserve 2 GiB of disk space (but no more than 1%) to allow the data disk to be replaced with a slightly smaller one in the future.
I removed the usage of pyparted because it was creating partitions that were smaller than requested, and despite significant effort, I was unable to resolve the issue. We had agreed to remove this library if it continued to cause trouble, and that moment has now arrived.