My son touched the Metadata 4x NVME mirror card in my Truenas Scale server while it was running. Need advice

I wasn’t paying attention and he reached into the top of the server while I was working on it. Now three out of the four Metadata drives (in mirror) have checksum errors.

I used the Storage configurator to remove each drive, one at a time and format/wipe it. Now I have one Metadata drive and three unassigned drives. I immediately tried to add the newly formatted drives back into the Metadata portion of the Pool but Truenas is insisting that the single remaining metadata NVME drive is a Stripe and refuses to add the three re-formatted drives back in to create the original, desired Mirror.

It explicitly errors out as a mismatch, Stripe - Mirror.

Any advice at all before I delete the entire array, re-make it, and restore it. Thank you.

Use 2 of the unassigned drives to create a new array, copy files from the existing drive to the new array, wipe the old array and add the remaining drives to the new array.

Output of zpool status from the cli?

Should be able to run zpool replace devname which is probably entirely numeric with the device path of one of the three new nvme

Thanks for the help ~ I am hoping we can avoid a pool destruction and restore.

` pool: bigstorage
state: ONLINE
scan: scrub repaired 0B in 04:26:45 with 0 errors on Sun Jun 11 10:26:48 2023
config:

    NAME                                      STATE     READ WRITE CKSUM
    bigstorage                                ONLINE       0     0     0
      raidz2-0                                ONLINE       0     0     0
        ed08ca0c-dafe-11eb-995e-a8a15941d663  ONLINE       0     0     0
        45ba7295-5f12-11ec-8c16-a8a15941d663  ONLINE       0     0     0
        9bae04cb-de08-11eb-a3fb-a8a15941d663  ONLINE       0     0     0
        03fa5731-da9d-11eb-99d1-a8a15941d663  ONLINE       0     0     0
        9dcbca8a-061d-11ed-8b31-a8a15941d663  ONLINE       0     0     0
        d5131533-db6a-11eb-8fb9-a8a15941d663  ONLINE       0     0     0
        298cf593-da30-11eb-9793-a8a15941d663  ONLINE       0     0     0
    special
      f04b6af1-d1b7-4d1d-aacc-e0a2b1cf20cf    ONLINE       0     0     0
    logs
      1bcd6efb-ca54-4ab9-8f0c-1a7ad9c07ea6    ONLINE       0     0     0
    cache
      fff38458-f2ea-4836-be88-eca43982660a    ONLINE       0     0     0
    spares
      79f496f1-80b8-4b87-8253-1d394afbf0b9    AVAIL   

errors: No known data errors

pool: boot-pool
state: ONLINE
status: Some supported and requested features are not enabled on the pool.
The pool can still be used, but some features are unavailable.
action: Enable all features using ‘zpool upgrade’. Once this is done,
the pool may no longer be accessible by software that does not support
the features. See zpool-features(7) for details.
scan: scrub repaired 0B in 00:00:29 with 0 errors on Wed Jun 14 03:45:30 2023
config:

    NAME        STATE     READ WRITE CKSUM
    boot-pool   ONLINE       0     0     0
      mirror-0  ONLINE       0     0     0
        sdj3    ONLINE       0     0     0
        sdl3    ONLINE       0     0     0

errors: No known data errors

pool: things_pool
state: ONLINE
scan: scrub repaired 0B in 00:00:00 with 0 errors on Sun Jun 11 05:00:01 2023
config:

    NAME                                      STATE     READ WRITE CKSUM
    things_pool                               ONLINE       0     0     0
      mirror-0                                ONLINE       0     0     0
        9490230e-0451-4fd8-9acc-1336548bf3b9  ONLINE       0     0     0
        c9b12567-4d67-41e2-93f4-656ac4a1715c  ONLINE       0     0     0

errors: No known data errors`

Attach not add?

Better to “replace” one dummy/missing drive listing though as the other drives were detached, instead if faulted, the 4 deep mirror is now a 1 deep raid0

Unless GUI is funky?

Personally I would be super careful now, with entire pool hanging on 1 drive, no redundancy

I would carefully DD the working drive To a formatted one, place copy to one side, then in truenas, attach one if the other formatted drives to the working one

1 Like

This sounds very interesting but it’s my understanding that the single remaining Metadata special vdev must remain associated with the Pool or the Pool will be destroyed. That is why I had quadruple redundancy.

1 Like

exactly. I’m sweating over here.

edit - Yes after I had done the thing, I realized I should have re-added each drive as it was formatted, and not removed all three.

1 Like

I haven’t used the GUI, but in ZFS, if you “attach” a drive, it becomes a mirror. If you “add” a drive, it becomes more like a raid0, kinda thing

1 Like

I’m looking into the syntax and usage for ‘attach’ r/n. thanks.

2 Likes

yeah, attach is what you want in this case. In the past when I’ve been in this situation the broken mirrors in special were still there? But they aren’t in your case. ‘attach’ and ‘special’ qualifiers… should… work in this case. Probably.

2 Likes

I want to thank you for everything that you do. I bought more than several 118GB Optane sticks from Newegg when your video dropped (at rock bottom prices), and also the pcie 4x NVME card for redundant Optane Metadata vdev on your recommendation.

As a dabbler for many years, your deep-dive insights are beyond educational and beyond appreciated. Also, “hi” to the whole gang.

1 Like

the re-silvering has begun. it is too late for the special vdevs to vote.

1 Like

whats the output of zpool status now?

1 Like

@Trooper_ish

sudo zpool attach -fw bigstorage f04b6af1-d1b7-4d1d-aacc-e0a2b1cf20cf nvme1n1

Everything worked. Thanks to all.

edit for @wendell

pool: bigstorage
state: ONLINE
scan: resilvered 844M in 00:00:20 with 0 errors on Thu Jun 15 21:41:45 2023
config:

    NAME                                      STATE     READ WRITE CKSUM
    bigstorage                                ONLINE       0     0     0
      raidz2-0                                ONLINE       0     0     0
        ed08ca0c-dafe-11eb-995e-a8a15941d663  ONLINE       0     0     0
        45ba7295-5f12-11ec-8c16-a8a15941d663  ONLINE       0     0     0
        9bae04cb-de08-11eb-a3fb-a8a15941d663  ONLINE       0     0     0
        03fa5731-da9d-11eb-99d1-a8a15941d663  ONLINE       0     0     0
        9dcbca8a-061d-11ed-8b31-a8a15941d663  ONLINE       0     0     0
        d5131533-db6a-11eb-8fb9-a8a15941d663  ONLINE       0     0     0
        298cf593-da30-11eb-9793-a8a15941d663  ONLINE       0     0     0
    special
      mirror-12                               ONLINE       0     0     0
        f04b6af1-d1b7-4d1d-aacc-e0a2b1cf20cf  ONLINE       0     0     0
        nvme1n1                               ONLINE       0     0     0
        b6d6f3e8-bd85-4591-bd41-bff2a31dca47  ONLINE       0     0     0
      0efdb1e7-949d-45c6-a869-9a8f1f87a152    ONLINE       0     0     0
    logs
      1bcd6efb-ca54-4ab9-8f0c-1a7ad9c07ea6    ONLINE       0     0     0
    cache
      fff38458-f2ea-4836-be88-eca43982660a    ONLINE       0     0     0
    spares
      79f496f1-80b8-4b87-8253-1d394afbf0b9    AVAIL   

errors: No known data errors

1 Like

OOoo that’s not good. That looks to me like a stripe of one drive and a 3 way mirror.

2 Likes

yes I used the Add GUI after re-creating the Special Mirror successfully, and it added the last drive by itself … oof.

So I corrected the single Metadata zaid0 situation … now 3-way mirror but the 4th drive is hanging out there on it’s own. Im 90% to the solution Im sure.

1 Like

welp, two options now. destroy pool or add one more nvme and have a striped mirror. You could possibly remove one of the mirrors from the 3 way mirror safely-ish. then you have a striped mirror which isn’t too bad in terms of redundancy.

1 Like

I actually have 2 unused 118G optanes on the motherboard, just hanging out. I’ll think this through. It’s a supermicro threadripper mobo.

1 Like

how big is that one that’s the one you need to add a mirror of that must be the same size (or larger)

then you’d have a striped mirror

1 Like

They are all identical 118G Optane sticks

edit for @wendell ~~

special
mirror-12 ONLINE 0 0 0
f04b6af1-d1b7-4d1d-aacc-e0a2b1cf20cf ONLINE 0 0 0
nvme1n1 ONLINE 0 0 0

      mirror-13                                                         ONLINE       0     0     0
        0efdb1e7-949d-45c6-a869-9a8f1f87a152  ONLINE       0     0     0
        962faaf1-1c99-4a4e-9eed-e54a2ce9cc02  ONLINE       0     0     0

This seems very unintended.

1 Like