Before just tearing the drive out of there, I manually selected the unit as "F--faulty" in the s/w raid manager. This caused the raid to become degraded. Then, I selected the unit as "R--hot remove". After this, I physically removed it. Nothing bad happened! Throughout this whole process, I had been playing a long audio file stored on an iscsi volume on this raid array. I did a unit rescan and found that, indeed, the drive was no longer there, but the other two were still there, and the array was fine. It was just degraded but still clean.
Then, I put the drive back in the slot and did a unit rescan. It showed up as "available"! So, I added it as a spare for MD0, which was in a degraded state. It immediately started rebuilding. Rebuilding is faster than initially building, for some reason (50MB/s vs. 20-40MB/s).
So, I'm still listening to the same audio file that I was at the beginning at this process without any hiccups or errors.
Tomorrow, I'm going to just remove the drive immediately without warning the software. We'll see how that goes!!!
I think the most important thing to do is to make sure AHCI is enabled. I'm using a 1U Supermicro SuperServer that has a 4-slot hot swap backplane and built-in SATA controller on the motherboard ("pretend" hardware RAID, aka BIOS-aware software RAID, but doesn't do much good for linux, since for linux it's both unnecessary and probably slower than linux's sw raid). I'll give you more precise information tomorrow.
We are working on the software RAID expansion this year. You can create another SW RAID and then add this to the existing Volume groups - poor man's expansion
RAID expansion would be nice, but we don't really require this in a software RAID.
Anyway, I just tried pulling a drive without warning the software ahead of time. It works great, with no interruptions (I was again streaming a media file while I pulled the drive). Right now, it is rebuilding. So, it seems like this software raid hot-swap works just as well as hardware raid.
Reply With Quote
