What happens if I ignore this error?

[zpoolChlorine]

Hi lads,

I have a janky build without ECC ram! It's a Sabertooth x79 with Intel(R) Xeon(R) CPU E5-2667 v2 @ 3.30GHz, 64GB RAM (GSKRILL, ddr3 unbuffered non-ecc). Running TrueNAS-12.0-STABLE off of a small SSD. Anyway all was fine until last night when a scrub triggered a critical error:

My questions are:

What terrible things will happen if I ignore it? Who are the possible culprits, is the hard drives, ram, or something else? It's not a file that is erroring out but a memory address? what is <0x18b6>:<0x2da2>?

 

[sretalla]

what is <0x18b6>:<0x2da2>?

It's a block (or some blocks) of metadata.

There's a possibility that some of your other data is already "missing" now as the metadata pointing at it is corrupt, so the filesystem has no way of locating it.

If you haven't already done that, take a backup of important data on that pool and recreate it.

Also have a look at the SMART data (after a long smart test on each) of your drives. That corruption suggests that probably 2 or even 3 of the disks have problems.

It wouldn't be impossible for it to be a controller error either, so consider what you're using (is it a Marvel SATA controller on that board?).

(and there it is... a Marvel RAID controller... https://dlcdnets.asus.com/pub/ASUS/mb/LGA2011/SABERTOOTH-X79/E8040_SABERTOOTH_X79_V3_WEB.pdf) That will be your problem... see here https://www.truenas.com/community/t...s-and-why-cant-i-use-a-raid-controller.81931/

Depending on how you had connected your disks, it seems only the Gray ports at the end of the SATA block are connected to the Marvel controller (item 10 on page 2-3), so maybe just swapping those over to the Intel ones would be enough (assuming you have enough ports in total).

 

[zpoolChlorine]

Depending on how you had connected your disks, it seems only the Gray ports at the end of the SATA block are connected to the Marvel controller (item 10 on page 2-3), so maybe just swapping those over to the Intel ones would be enough (assuming you have enough ports in total).

wow thanks! I'll swap the ports around, had no idea about marvel... I'll run a bunch of scrub tests and smart tests to verify.

 

[sretalla]

I'll run a bunch of scrub tests and smart tests to verify.

You may still find that corruption will persist, so you would need to rebuild the pool if that's the case, but with all disks on the Intel controller, you should be able to avoid corruption in future after that.

 

[zpoolChlorine]

You may still find that corruption will persist, so you would need to rebuild the pool if that's the case, but with all disks on the Intel controller, you should be able to avoid corruption in future after that.

Ya I rewired the ports and rebuilt the pool from scratch. No issues on multiple scrubs cheers