Adding new disk to server results in "failed to import pool 'rpool'"

[keson]

Dear Proxmox gurus,
I am here new. Spend several days trying to resolve my issue by looking through this forum, but not sure I saw similar situation, although the symptoms seem to be discusses a lot here. I would need to hear some hints as I am fairly new in this and need to solve a situation in production environment.

Installed Proxmox (latest version as of now) on HP GL380 several months ago and all works just so excellent. I have 4 HDDs for data and virtual machines and ONE SSD drive for ZIL logs and cache. The HDDs are in Mirror, the SSD is single. It gets 7% wear-out and I wanted to add second SSD disk for the LOGS and CACHE to make it a mirror for security.

The problem is following:

As soon as I insert the new disk in the server and I boot it up (as it can't be added on the fly) the Proxmox fails to load with an error we can see in many local posts:

Failed to import pool 'rpool'
Manually import the pool and exit

Now the question is HOW to properly add the new disk to a server to avoid such situation or better to ask, whether this is a default behaviour whenever a new disk is inserted to a server?

As soon as I remove the disk, I can start the server again.

As I am afraid to do any trial error experiment as it is a production environment and I have no other system to play with, I would like to ask how to proceed.

Basically my aim was to insert the new disk, then delete the current CACHE and LOGS, create a new mirror from disk SDC and SDF and again use it for CACHE and LOGS... Is that feasible?

This is my current configuration:

(see zpool status)

RPOOL
SDA2 + SDB2 = MIRROR-0
SDD + SDE = MIRROR-1
LOGS SDC1
CACHE SDC1

(see zpool list)

The new disk, when I add it would be SDF I expect.

I apologise for this novice question, but I rather want to make sure what I do before I drive to the location and do the new disk insertion and restart on the spot.

Happy xmas!

 

[Rob Loan]

Sounds like the new disk has been part of an rpool in its past life. One of there ways:

1) figure out a way to plug in the new disk while pve is running.

2) erase the front of the disk

3) in single user mode on the pve failed boot, zpool import by ID, and exit to continue the boot

 

[joshin]

Also make sure you're using the hard-coded disk ID, and not the /dev/sd? drive assignment.

 

[keson]

Thanks Rob and Joshin,
You pretty much proposed what I was intending to do. I was really surprised, that the HP Proliant GL380 does not allow to insert a hard drive while the server runs. Perhaps it is because it is a SATA and not SAS/SCSI disk... But when I do insert it AFTER the POST initialisation, the disk is NOT recognised so when I try to access the disk later (fdisk -l etc.), it is "not" there so to say.

The new SSD disk I try to add is a disk I got from a system, where it was used as a windows disk for a week or so. I am waiting for a delivery of brand new two professional SSD drives, but before they come, I would like to improve the reliability of current setup by adding this second used SSD drive. What I am glad is that you identified the potential cause - the disk was used before... Hope it is the cause.

I am perhaps much failing in the second step too. I did try to delete all the partitions in many systems - under windows, under linux, under mac. Completely deleted all partitions etc. fdisk seems to be the right way, I believe, is there any recommended way of erasing the disk?

If I understand the third step correctly, I will boot in the same way as I did before with the error and then I do manually IMPORT the new disk by:

zpool import 12345678901234567890

correct?

then I should remove the current cache and log:

zpool remove rpool /dev/sdc1
zpool remove rpool /dev/sdc2

Then I should create a mirror from the newly added ssd disk (hopefully sdf) and the old ssd disk (sdc) and lastly create a new cache and log...

zpool add rpool log mirror /dev/sdc1 /dev/sdf1
zpool add rpool cache /dev/sdc2 /dev/sdf2

So basically all I need to do is to make sure that I can "see" the new disk ID, and then zpoll shall be able to use it, right?

I am just wondering, if this is the "right way" of doing it or just some kind of workaround. I do want to do things the right way so I would love to learn how to It is hard to change the mind after 30+ years with classic filesystems, this ZFS for me is simply so new.