I'm encountering some errors while trying to add an SSD OSD to my Proxmox Bluestore Ceph setup. The logs show several stderr messages that mention an "unparsable uuid" and an issue with the min_alloc_size not being power of 2 aligned, which leads to a failure when executing the mkfs command. I've already tried identifying the disk by using commands like wipefs and sgdisk to zap it. My main question is whether these issues indicate a problem with the SSD itself, and if replacing it is necessary. Any insights?
2 Answers
Before you assume the SSD is the issue, have you checked for any I/O errors in the host's journal? The parsable uuid error definitely stands out. You might find more clues in the logs before this error occurred. Getting to the root of that first could help clarify the problem.
Good idea! You might find something that points to a different issue altogether.
You shouldn't have to replace your SSD. The messages you're seeing suggest that the drive is fine, but the min_alloc_size you're trying to use is what's causing the failure. It's not set to a power of 2, which is why you're getting that invalid argument error. Just ensure you wipe the disk properly with ceph-volume, and check that your Ceph config doesn’t have a wrong bluestore_min_alloc_size setting. Fixing that should let you add the OSD without issues.
I checked the config and saw that the bluestore_min_alloc_size was set to 16000. Should I just change it to 4096 like you suggested?
Yes, setting it to 4096 should work! If you do that and wipe the disk correctly, you should be all set.
Thanks for the tip! I’ll dig into the task logs for anything suspicious before that error.