USB drive degraded used as datastore cause Proxmox crash

[gannick]

Hello everyone,

I installed PBS on the same physical server as my PVE.

Since 10 days, my PVE is crashing and I must make hard reboot to make it alive again.

Today, crashed again, I decided to investigate. I was looking after the internal disks (2 SSD ! SMART tests are passed without problems).

After that, I look on 2 USB disks connected to my server. One of them is used as a datastore for PBS. ZFS is used on a 4 TB USB disk. This disk appears as "DEGRADED".



My problem is that I would like to stop my PBS instance and I cannot see how to do it.

What is the right procedure for ejecting that drive ?

What I plan to do (but I would like to have expert advice) :
1) Stop PBS (how ?)
2) Make backup job as inactive from PVE
3) Delete the datastore used by PBS
4) Eject the USB disk


Any advice may be great ! Thanks by advance.

 

[gannick]

Or do you know how I could replace my datastore which is located on the USB disk which is going bad ? May I delete the datastore and create a new datastore with the same name on a new disk ?

 

[dcsapak]

HI,

first, i don't think it's a good idea to run a datastore on zfs with a single disk via usb, because
no redundancy, so in case something goes wrong the pool could be broken
usb is sadly often unreliable, be it the controllers, cables or sata adapters, so it's more likely to fail in the long term

1) Stop PBS (how ?)

you stop the 2 main services:

systemctl stop proxmox-backup proxmox-backup-proxy

since it's a zfs pool you want to 'export' it before e.g. wiping the disk (see 'man zpool-export')

but in any case, I'd rather:
* look into why the pool is degraded (e.g. faulty disk/usb cable/controller/etc) and fix that
* use a different setup (like said above, more disks, no usb, etc.)

 

[gannick]

HI,

first, i don't think it's a good idea to run a datastore on zfs with a single disk via usb, because
no redundancy, so in case something goes wrong the pool could be broken
usb is sadly often unreliable, be it the controllers, cables or sata adapters, so it's more likely to fail in the long term


you stop the 2 main services:

systemctl stop proxmox-backup proxmox-backup-proxy

since it's a zfs pool you want to 'export' it before e.g. wiping the disk (see 'man zpool-export')

but in any case, I'd rather:
* look into why the pool is degraded (e.g. faulty disk/usb cable/controller/etc) and fix that
* use a different setup (like said above, more disks, no usb, etc.)

Hi @dcsapak,

Thanks for your answer. I will try.

I found that about removing a datastore. What do you think about ?

For my setup which is not very smart, I agree. But my Proxmox PVE and PBS are working on a small computer (such as an Intel NUC). So I can only set up 2 NVME disks. It is a server for a domestic use... That's why I used some USB disk, which is not a so kind idea...