Samsung PM983a Enterprise NVME: Failed to flush random seed file: Time out when using ZFS boot

A

anime

New Member
Apr 2, 2024
15
1
3
Hi all,

I'm dumbfounded by this issue than only manifest when I'm using this particular drive as ZFS RAID0 for PVE host boot. Installation of PVE host goes smoothly (secure boot turned off) but upon reboot I'm presented with red text saying the following and takes me back to BIOS screen:

Failed to flush random seed file: Time out
Error opening root path: Time out

This drive is an enterprise M.2 22110 and is fully healthy, I can boot PVE host off of it if I keep the filesystem as EXT.

I also know that non-enterprise NVME plays well with ZFS as boot disk for PVE host. So there must be something along the encryption line on these enterpise causing this issue.

Anybody experienced anything similar or shed some light on possible workaround?

Thanks a lot.
 
Last edited:
Adding additional information if that is of any help to resolve this issue. Above error was with PVE 8 during boot.

With PVE 7, I still get similar error "Failed to flush random seed file: C0798...."

Any help is much appreciated.
 
I don't recognize the error, but I use these drives in all my hosts without any issues. Maybe a firmware update would we worth giving a try?
 
  • Like
Reactions: anime
dignus said:
I don't recognize the error, but I use these drives in all my hosts without any issues. Maybe a firmware update would we worth giving a try?
Click to expand...

Thanks for the insight, definitely worth a try.

How do you go about getting the latest firmware for enterprise SDD? I would imagine Samsung Magician won't work with these drives.
 
anime said:
Hi all,

I'm dumbfounded by this issue than only manifest when I'm using this particular drive as ZFS RAID0 for PVE host boot. Installation of PVE host goes smoothly (secure boot turned off) but upon reboot I'm presented with red text saying the following and takes me back to BIOS screen:

Failed to flush random seed file: Time out
Error opening root path: Time out

This drive is an enterprise M.2 22110 and is fully healthy, I can boot PVE host off of it if I keep the filesystem as EXT.

I also know that non-enterprise NVME plays well with ZFS as boot disk for PVE host. So there must be something along the encryption line on these enterpise causing this issue.

Anybody experienced anything similar or shed some light on possible workaround?

Thanks a lot.
Click to expand...
I'm having a similar issue with my 2 new Samsung PM983 drives. Can't seem to figure it out...
 
  • Like
Reactions: anime
B.Otto said:
It has been a while, but back when we set up our system with ZFS-boot on SSDs, we had to configure grub to delay the ZFS invocation. Maybe you need to add something like rootdelay=30 to grub-cmd:
https://pve.proxmox.com/wiki/ZFS:_Tips_and_Tricks#Boot_fails_and_goes_into_busybox
Click to expand...

Thanks for the pointer. I'm using systemd-boot not GRUB, have to update as following:

Code: 
edit /etc/default/zfs, set ZFS_INITRD_PRE_MOUNTROOT_SLEEP='4', and then issue a "update-initramfs -k 4.2.6-1-pve -u"

Because I'm stuck at boot with the error mentioned above and don't have access to the prompt, how do I go about editing and updating the boot image?
 
anime said:
how do I go about editing and updating the boot image?
Click to expand...
I think I am stuck near where you are. I was able to do the editing after running the following command:

zpool import -R /rpool -N rpool

Now I am stuck at the next step. The command in the documentation is specific to one kernel version. If you want to apply this change to all, which I think is the correct way to do it, you use the following command:

update-initramfs -k all -u

However, this command fails with the following error:

sh: update-initramfs: not found

So I am still stuck unable to boot.

Side note: ZFS_INITRD_PRE_MOUNTROOT_SLEEP doesn't exist in /etc/default/zfs. I had to append the key/value pair to the end of that file. Because I am unable to perform the next step, I don't know if this config change works or not.
 
Last edited:
  • Like
Reactions: anime
utkonos said:
I think I am stuck near where you are. I was able to do the editing after running the following command:

zpool import -R /rpool -N rpool

Now I am stuck at the next step. The command in the documentation is specific to one kernel version. If you want to apply this change to all, which I think is the correct way to do it, you use the following command:

update-initramfs -k all -u

However, this command fails with the following error:

sh: update-initramfs: not found

So I am still stuck unable to boot.

Side note: ZFS_INITRD_PRE_MOUNTROOT_SLEEP doesn't exist in /etc/default/zfs. I had to append the key/value pair to the end of that file. Because I am unable to perform the next step, I don't know if this config change works or not.
Click to expand...

Right. Apparently, for systemd boot I could hit "e" to make temporary non-sticky change to add
Code: 
rootdelay=30
and hit "Enter" to no avail. I still throws the same error.

Too bad I have to use an extra NVME as boot disk just to circumvent this issue. Would love to hear if someone out there have figured how to get away without it.
 
I have scrapped trying to get ZFS working for the moment. I was able to install with zero issues using BTRFS. Unfortunately, it is a tech preview, so I am unsure if this is a configuration I want to use.
 
Has anyone figured out how to resolve this besides simply using a filesystem besides ZFS? I have the same error trying to install proxmox in a ZFS Raid01 mirror using two Samsung PM983 22110 enterprise m.2 nvme drives.

Failed to flush random seed file: Time out
Error opening root path: Time out
 
Last edited:
After retrying the proxmox install again with the same ZFS RAID1 mirror on the same nvme drives it worked. Sorry I don't have a better solution.
 
  • Like
Reactions: anime
I got the same issue with my 2x 983 DCT drives. Tried to install with zfs RAID 1 multiple times but I always get the same error.

Installed with BTRFS RAID1 for now.
 
@vfedetz glad to hear that. Can you please confirm what FW is your PM983 running?

nvme list

Is it factory loaded EDB7602Q or latest EDB7902Q?
 
ED1Zx said:
I got the same issue with my 2x 983 DCT drives. Tried to install with zfs RAID 1 multiple times but I always get the same error.

Installed with BTRFS RAID1 for now.
Click to expand...

What FW are you running on your 983? EDB7602Q or latest EDB7902Q (latest)?

nvme list

I'm curious if latest FW can help address the ZFS RAID 1 boot issue. Instructions to update FW is here: https://samily.it/posts/fw-update-pm983/

And the latest FW bin file can be downloaded from here: https://files.hddguru.com/download/Firmware updates/Samsung/SSD DRIVES/PM983/
 
Download Samsung DC Toolkit:
How to Use:

  1. Download and install Samsung DC Toolkit.
  2. Open Command Prompt and run:
    bash
    Code: 
    samsung-dctoolkit
  3. Check the connected SSDs and select the one you want to update.
  4. Follow the on-screen instructions to complete the firmware update.
Important Notes:

  • Do not turn off or reboot your system during the update.
  • Backup your important data before proceeding.
Additional Information:
 
shbaek said:
Download Samsung DC Toolkit:
How to Use:

  1. Download and install Samsung DC Toolkit.
  2. Open Command Prompt and run:
    bash
    Code: 
    samsung-dctoolkit
  3. Check the connected SSDs and select the one you want to update.
  4. Follow the on-screen instructions to complete the firmware update.
Important Notes:

  • Do not turn off or reboot your system during the update.
  • Backup your important data before proceeding.
Additional Information:
Click to expand...

Are these instructions just to update the FW or does it help address problem at hand?

Is there any benefit to using samsung-dctoolkit over nvme-cli for FW update? In both cases you need a .bin file.
 
anime said:
Are these instructions just to update the FW or does it help address problem at hand?

Is there any benefit to using samsung-dctoolkit over nvme-cli for FW update? In both cases you need a .bin file.
Click to expand...
Samsung DC Toolkit is primarily for firmware updates, but updating firmware can sometimes resolve compatibility issues.

The key difference between samsung-dctoolkit and nvme-cli is that:
  • Samsung DC Toolkit is Samsung’s official tool, ensuring compatibility and automatic firmware verification.
  • nvme-cli is a generic NVMe tool that requires manual commands and may not support proprietary Samsung firmware checks.
If your issue is firmware-related, Samsung DC Toolkit is the recommended choice. If you need more control, nvme-cli can be used with the correct .bin file.
 
  • Like
Reactions: anime
You must log in or register to reply here.
Top Bottom