I have a server with a raid controller that I cannot put in to IT mode, at least that I have been able to find.
I have set up all my drives as single drive raids in the Raid card so that Truenas Scale can see them. Can I swap a raid controller with an HSB in IT mode without rebuilding array?
I have an Intel S2600GZ with an RMS25CB080 Raid Card.
There is chance for failure as truenas may not be able to identify the drives without the raid card passing them through.
Canāt think of a smart way to confirm that doesnāt put pool at risk - the dumb & risky way would be to plug one of the drives from the raid card into the motherboard & see if truenas still identifies it as part of the pool.
Why is this dumb? Because if it doesnāt & you try to undo these changes, will the raid card present that drive in the same way that truenas is expecting again? Does the put you at risk with redundancy? etc.
Iād plan a backup strategy just in case. In fact Iād recommend one even if you donāt want to get an LSI HBA (or equivalent).
Thatās what I was thinking of doing.
My raid card is directly stacked on the mobo like a pi hat kinda deal and then plugged into an expander. I was thinking I would pull one of the banks which has only my NVR drive on it and put it direct to the mobo and see what happens. No important data there.
If it worked, I was going to see about just plugging the expander into the mobo and see how it went.
TrueNAS just imports the pool via ZFS, ZFS doesnāt care about anything other than itās labels that are a part of itās partition (UUID+ZFS label). if the goal is to see if the disk is a part of a single raid volume, you might just see regular disks, you might not, it depends on the implementationā¦ Intel RST slices will just look like single ānormalā disks outside of RST (no idea about the hardware controller described though)
ZFS (itself) isnāt like Unraid where if you use a different enclosure and now the disk has a different name or serial, your disk is assumed to be new/unassigned. ZFS also isnāt like HW raid which identifies generally by WWID or other hardware characteristics. ZFS will generally ājust import fineā as long as the disk is working.
regarding OPs questions, I would have doubts, Iāve never seen anyone use a hardware Intel RAID card. but if you have options for disabling things like hardware caching, enabling SMART, and anything you can to stop the RAID card from doing things that an HBA wouldnāt do. and of course do regular scrubs and assume that using unknown hardware is going to possibly have unkown issues we canāt possibly warn you about.
So youāre saying that my idea is less dumb & risky & should work?
It does look like because the raid controller set each drive up as a virtual disk/volume that simply bypassing them and plugging directly into the mobo did not workā¦ though the bios could still see the proxmox install, it just simply wouldnt boot.
I would have to rebuild everything. I have everything backed up, I just didnt want to rebuild after I justā¦ rebuilt.
it should be possible to āreplaceā every disk one at a time.
as long as you are comfortable running the pool degraded for a bit it will be the easiest way, and you wonāt need any spares. (assuming itās not a stripe pool)
if you have a spare disk lying around, you can also do an āonlineā replacement, which will maintain redundancy during the resilver (you would just click replace on a disk that isnāt degraded/removed yet)
This might indeed be the smartest way - assuming he has a free PCIE slot because the S2600GZ is possibly the wildest motherboard Iāve ever seen.
