Corrupt files, system panic

[florisvb]

I had a series of problems, including "welcome to nginx" gui, and somewhere along the way a lot of corrupt files. I upgraded to 8.3.1, replaced the corrupt files with fresh copies and ran a scrub, but still had 4 files with permanent errors, with names like:

home/Floris:<0x1f0d>

One of them was in my Jail, which I'm not using. I thought maybe if I delete that, that error will go away, and if that worked I could try the same with my other dataset. But after trying to destroy the Jail (via the webgui) I now get this when I start up the machine:

panic: Solaris(panic): zfs: allocating allocated segment (offset=2284353604608 size=2048)

Any ideas on how to solve this? Thanks!

- Floris


System specs:

CPU / Motherboard: ASUS C60M1-I AMD Fusion APU C-60 (1.0GHz, dual core) AMD Hudson M1 Mini ITX Motherboard/CPU Combo
8 GB RAM
4 GB USB stick
Intel GigE Network card
4x 2 TB WD hard drives

 

[florisvb]

Turns out my RAM was bad, and that seems to have permanently destroyed the data on my drives. I tried a fresh install of FreeNAS (8.3.1), which starts up fine, but then as soon as I either upload my old config file, or try an auto-import of volumes, I get the same panic.

 

[cyberjock]

Yeah. Bad RAM is basically instant permanent fail for ZFS. This kind of example shows why Sun and Oracle recommend ZFS only for ECC RAM systems.

The thing to take away from this is ZFS is designed to repair corruption from any hardware or software issue... except bad RAM.

Bad RAM is a real killer, especially if you think about the fact that you should be doing scrubs regularly, so not only is your new data potentially corrupt, but you potentially corrupted the entire zpool with garbage because of bad RAM. Just not a good place to be in. :(