[SOLVED] Issues with PVE 5.0 installer when there is a Digital Devices Cine S2 v6.5 DVBS card

[m00n]

Hi,

maybe the 5.0VE installer has a showstopper bug, when installing to a ZFS Raid-1 mirror set with 2 SSDs. The Installer stopps with an error msg. during creating the partitions.
The error msg. is: unable to create swap space.
I have no clue, whats wrong. Before I start a BR I wanted to ask here before.

With 4.4VE installation I had no problems. The installer created the mirror set without any problems.

Step by step:

HW: Supermicro X10SLH-F, Xeon 1275LV3 (socket 1150), 32GB ECC, 2x Samsung 850 PRO SSD (256MB each).

Proxmox ISO-file: 5.0-af4267bf-4.iso
ISO via. dd to a Sandisk USB Stick (16GB). Booting works without problems. After booting from USB the Proxmox installer showed up.

1) Under opions I choosed "filesystem zfs RAID 1"
2) The installer showed 3 entries:

1. Harddisk 0 /dev/sda (238GB Samsung SSD)
2. Harddisk 1 /dev/sdb (238GB Samsung SSD)
3. Harddisk 2 /dev/sdc (14GB, Extreme)

3) I changed Harddisk 2 to: -- do not use --

(These settings worked with Proxmox VE 4.4 without any problems.)

4) I did not changed anything under Advanced options.
5) Now following the Installer and entering Password, IP-Settings, etc.
6) When the installer begins to create partitions the installer stopped with "Installation failed"
7) Dialog-Box opend: "unable to create swap space"
8) Switching to console with CTRL+ALT+F2: see picture

In another attemp, I tried to remove all partitions on both SSDs before setup. But same result.

I dont want to create everything manually so I switched back to 4.4VE.

Anyone has any idea? Maybe a bug in the installer?

Thx, kai


Console output after the installer stopped with an error:

 

[Rhinox]

There was problem with using disks which have been previously used somehow (and even cleaning with dd did not help), but I thought this had been fixed in PVE 5 "final" already...

 

[fabian]

please boot in debug mode, attempt the install and save the complete debug log (available in /tmp/install.log in the debug shell).

 

[m00n]

@Rhinox
I only removed all partitions and created a new GPT partition table via gparted on both drives. But these steps didn't solve the issue. I didn't try to overwrite with dd.

@fabian
I followed your instructions and uploaded the saved install.log file from the debug mode setup. Before booting the proxmox iso I manually deleted all partitions from both SSDs. So both drives were empty before starting the installer.
Setup options again filesystem zfs RAID 1

 

[fabian]

looks like the device node for the zvol is not created (fast enough?). could you check whether it exists in the debug shell?

zfs list rpool/swap
ls -lh /dev/zvol/rpool/swap

if it does not, could you execute the following and then check again?

udevadm trigger --subsystem-match block
udevadm settle --timeout 10

 

[m00n]

looks like the device node for the zvol is not created (fast enough?). could you check whether it exists in the debug shell?

zfs list rpool/swap
ls -lh /dev/zvol/rpool/swap

if it does not, could you execute the following and then check again?

udevadm trigger --subsystem-match block
udevadm settle --timeout 10

Screenshot with output attached. There is no /dev/zvol.

 

[Rhinox]

Try to repeat all those installer-commands manually, in debug-console (everything that starts with "#" in your install.log). Or at least try that "zpool create -f ..." command. Maybe some error shows. You might also try to increase that udevadm-timeout...

 

[m00n]

Try to repeat all those installer-commands manually, in debug-console (everything that starts with "#" in your install.log). Or at least try that "zpool create -f ..." command. Maybe some error shows. You might also try to increase that udevadm-timeout...

Repeating all installer-commands from the beginning makes sense? As fair as I can see, the rpool was created correctly and there is a Raid-1 mirror. Also an 8.5G swap partition rpool/swap was created. But there is no device node /dev/zvol. So mkswap failed when executed, because the swap space couldn't be mounted. I tried to increase the timeout, but it doesn't change, still no device node. I have no clue what's going on.

 

[m00n]

@Rhinox
Ok, I manually erased all partitions and booted into debug-mode. I switched over to the console (CTRL+ALT+F1). Then I executed all installer-commands from the instal.log manually (lines beginning with #). I increased udevadm-timeout to 60 with no effect. All commands worked without errors. Only the last "mkswap -f /dev/zvol/rpool/swap" command failed, because of the missing device-node /dev/zvol. So we need another approach to solve the problem.

 

[Rhinox]

...Only the last "mkswap -f /dev/zvol/rpool/swap" command failed, because of the missing device-node /dev/zvol. So we need another approach to solve the problem.

/dev/zvol is directory, device node is elsewhere. But may I suggest you try to create it manually, with "mknod"?

FYI in my installation:
/dev/zvol is drwxr-xr-x root:root
/dev/zvol/rpool is drwxr-xr-x root:root

and in /dev/zvol/rpool there is link:
lrwxrwxrwx root:root swap -> ../../zd0

device-node /dev/zd0 is:
brw-rw---- root:disk 230,0

so you'd have to do something like:
mknod /dev/zd0 b 230 0
(and then use chgrp & chmod to fix group and access rights)

 

[m00n]

/dev/zvol is directory, device node is elsewhere. But may I suggest you try to create it manually, with "mknod"?

FYI in my installation:
/dev/zvol is drwxr-xr-x root:root
/dev/zvol/rpool is drwxr-xr-x root:root

and in /dev/zvol/rpool there is link:
lrwxrwxrwx root:root swap -> ../../zd0

device-node /dev/zd0 is:
brw-rw---- root:disk 230,0

so you'd have to do something like:
mknod /dev/zd0 b 230 0
(and then use chgrp & chmod to fix group and access rights)

Ofc, you're right, /dev/zvol and /dev/zvol/rpool are directories.

mknod is not necessary, because the node /dev/zd0 is existing.

This is what I have done now:

1) After the installation aborted, I switched to console:

/dev/zd0

was created.

2) Creating 2 directories manually:

mkdir -p /dev/zvol/rpool

3) swap link creation in /dev/zvol/rpool:

ln -s /dev/zd0 /dev/zvol/rpool/swap

4) Setting up swap area:

mkswap -f /dev/zvol/rpool/swap

With these steps above, the mkswap command from the install.log worked this time. But now I can't continue, because the installation was aborted before.

The ISO-Installer has to be fixed. I will go and file a bug-report now. Thx.

Curious, that no bug-report already exists and that nobody noticed this issue with Proxmox 5 til today.

 

[m00n]

Bug-Report
https://bugzilla.proxmox.com/show_bug.cgi?id=1463

 

[m00n]

The problem is fixed let's say with a workaound.
For more information please take a look in the bug-reports here:

https://bugzilla.proxmox.com/show_bug.cgi?id=1463#c2