cjpeckover

Sorry if this has been brought up before, but I'm currently having an issue with my pcie Coral TPU constantly being dropped by Unraid. Every few minutes the Coral TPU Driver dashboard will display the status, temp, frequency as SHUTDOWN, which in turn will cause Frigate to complete stop functioning since the TPU can't be found (and fallbacks aren't yet an option). I'd found another thread that suggested monitoring Coral temperatures with script, but nothing out of the ordinary from that. The temperature will hover around 38-42C, then when the TPU dies it'll just report -89.70 until either Unraid is completely rebooted (full shutdown) or the pcie slot is rebooted with a script. Have there been reported instances of coral pcie TPUs malfunctioning after a couple years of use? Any help, or pointing to another thread, would be appreciated.

Right, fair enough. Thanks all for your help!

Ah right, that's a good point. I noticed in the docs they mention: However, I can't find anything on how this could be done, especially with what you said about Unraid likely making some writes while just mounting/starting the array.

So in this case I could start in maintenance mode, and then run the default mount commands Unraid uses during a normal array start, with -r flag? Looks like for the array disks it's usually something like: mkdir -p /mnt/disk1 mount -t xfs -o noatime,nouuid /dev/md1 /mnt/disk1 The caches are a bit different. Think I'd be fine then mounting the disks this way, read only, and the caches normally?

If I kept the array running and no writes were done (hypothetically), would this still technically be the case?

I'm upgrading my parity drive to a larger capacity and am wondering if it's possible to rebuild a data drive with the old parity in the case that a drive failed during the parity rebuild. For example, if I swap in the new drive and start the parity rebuild and a data drive failed, could I swap in the old parity drive and a new data drive, check that the parity is valid, and then start rebuilding the failed drive? The reason I ask is I'm trying to determine the best way to upgrade the parity while keeping the array protected. The best I've seen so far is to assign the new parity drive as parity #2, rebuild, then unassign parity #1. I'd rather not deal with new configs setting this up though if I can just swap them out and be safe with the old one should something catastrophic happen during rebuild. Thanks for any help!