Avago 3108: rebuild degraded hardware RAID5?

kindzmarauli

8/29/24, 10:08 PM

Supermicro TwinPro chassis with dual SYS-1029TP-DC1R server modules, each with Supermicro X11DPT-PS motherboard and Avago 3108 RAID card.

OS: VMware ESXi 6.7.

Problem: can't figure out how rebuild an existing RAID5 array after a "failed" drive was replaced. (Not really failed; rather, it's an attempt to replace spinning disks with SSDs one-by-one, rebuilding the RAID group every time - as opposed to rebuilding the RAID group from scratch.)

The "failed" drive: Seagate 7E2000 1TB mechanical drive (ST1000NX0303), 4K sector size
The replacement one: Micron 5400 Max SSD 1.92TB (Micron_5400MTFD), 512B sector size

(Is it possible this RAID controller can't mix drives with different sector sizes in the same volume aka "virtual drive"?)

Notes:

The controller is configured to start rebuilding on its own if a hot spare is available, or if a "UG" (unconfigured good) drive is inserted. (No rebuilding was happening in this case.)
Setting the replacement drive as a "global hot spare" - no change, no rebuilding in progress, no "rebuild the array" option anywhere in the UI.

Thanks for any ideas!

(There are more screenshots available of various UI parts.)

1 + 2

lsi

hardware-raid

raid5

failed

Nikita Kipriyanov

8/30/24, 5:44 PM

It is almost certainly the controller can't mix different sector sizes, but also in LSI controllers there is a knob which enables the mix of HDD and SSD in the same array. It is obvious it needs to be "enabled" in your case. Also, I vote to close this as this doesn't show a reasonable management practice; if you are managing RAID in a business environment it's reasonable to have a stock of replacement drives (cold spares) as well as hot spares connected to the backplane; *all of same kind*.

kindzmarauli

8/30/24, 6:15 PM

we do have stacks of spare drives of the same kind; it would be a good practice before voting for closing anything or implying poor management practices to check if the situation allows for hot spares (it doesn't, not with only 4 available drive slots). This particular question and problem were a direct result of a sanctioned experiment where the existing spinning disks would be replaced one-by-one with SSDs; the host in question was taken offline (put in "maintenance" before performing any actions described above). I vote to qualify your comment and voting action as a moderation overreach. ;)

Score:1

Server

kindzmarauli

8/29/24, 10:52 PM

I can't be 100% sure that's what it is - yet suspect the lack of "rebuild" option, and the reason the RAID5 volume doesn't start a rebuild on its own, is that an existing volume with 4Kn drives (4K sector size) can't take drives with a different sector size (512B). Wish the UI would do a better job indicating that.

What happened next:

The moment I put the original drive back in, it got marked as "UB" ("Unconfirmed Bad")
I then marked it "Unconfirmed Good" via "PD Mgmt" UI.
The drive immediately became "foreign".
I then "imported foreign configuration" in "VD Mgmt" UI.
The drive then became a member of the original volume / drive group, and the volume started rebuilding. It's currently at 15% after about 20 minutes. The LED is solid red and the blue LEDs on all drives are blinking.

(I'll mark my own answer as the "solution" for now - yet if there comes a long a better answer, will be happy to switch to it as the answer.)

+ 0

Elon Musk

I sit in a Tesla and translated this thread with Ai:

EN: Avago 3108: rebuild degraded hardware RAID5?

Post an answer

Most people don’t grasp that asking a lot of questions unlocks learning and improves interpersonal bonding. In Alison’s studies, for example, though people could accurately recall how many questions had been asked in their conversations, they didn’t intuit the link between questions and liking. Across four studies, in which participants were engaged in conversations themselves or read transcripts of others’ conversations, people tended not to realize that question asking would influence—or had influenced—the level of amity between the conversationalists.