Is there an ISO that supports selecting HWE?

[kazuki]

I was previously running this on bare metal on a PowerEdge R650. Since I needed more performance, I replaced it with a PowerEdge R7725 with dual EPYC 9965.
However, the currently distributed ISO seems to freeze while booting the Ubuntu installer. After selecting the bare metal install option in GRUB, it just hangs on a black screen.
Interestingly, if I use the ISO from the official Ubuntu site and select HWE, it installs without any problems.
Has anyone else run into this, or is there an ISO with HWE available?

Also, I installed the eve-ng pro pkg, but when I try to launch the HTML5 desktop or Integrated Wireshark, I get a “Disconnected” message in Guacamole. Not sure if it's related though.

[rusty725]

You can use legacy installation and select hwe. After you install ubuntu then install Eve.https://www.eve-ng.net/wp-content/uploa ... 5-2025.pdf page 33.

[kazuki]

Yeah, I did follow the cookbook for the legacy installation.
But anything that uses Guacamole shows a "Disconnected" message (e.g., Wireshark or the HTML5 desktop).
So far, that’s the only issue I’m seeing.

unl_wrapper.txt log

Error: No such object: html5Desktop-admin
Mar 09 15:46:21 ERROR: create docker html5Desktopadmin
Mar 09 15:46:21 INFO: started process is
Mar 09 15:46:21 INFO: starting ip link add docker0_0 type veth peer name screen_0
Mar 09 15:46:21 INFO: starting ip link set dev screen_0 up
Mar 09 15:46:21 INFO: starting brctl addif docker0 screen_0
Mar 09 15:46:21 INFO: starting docker -H=tcp://127.0.0.1:4243 inspect --format "{{ .State.Pid }}" html5Desktop-admin
Mar 09 15:46:21 INFO: starting ip link set netns 3867304 docker0_0 name eth0 address 50:00:00:00:00:00 up
Mar 09 15:46:21 INFO: starting /opt/unetlab/wrappers/nsenter -t 3867304 -n ip addr add 172.17.136.141/16 dev eth0
Mar 09 15:46:21 INFO: starting /opt/unetlab/wrappers/nsenter -t 3867304 -n ip route add default via 172.17.0.1 dev eth0
Mar 09 15:46:21 replace into guacamole_connection_parameter ( connection_id , parameter_name , parameter_value ) values ( 34957,'resize-method','reconnect' );
Mar 09 15:46:21 Mar 09 15:46:21 DEBUG: /html5/#/client/MzQ5NTcAYwBteXNxbA==?token=2D136972EB3A5223A2C95FC444BB433FC63E5A783A1DEDA7DC3FEC5DF8280474
Mar 09 15:46:23 Mar 09 15:46:23 Online Check state: Valid
Mar 09 15:46:23 INFO: started process istrue
Mar 09 15:46:23 INFO: starting ip link add docker0_0 type veth peer name screen_0
RTNETLINK answers: File exists
Mar 09 15:46:23 INFO: starting ip link set dev screen_0 up
Mar 09 15:46:23 INFO: starting brctl addif docker0 screen_0
device screen_0 is already a member of a bridge; can't add it to bridge docker0.
Mar 09 15:46:23 INFO: starting docker -H=tcp://127.0.0.1:4243 inspect --format "{{ .State.Pid }}" html5Desktop-admin
Mar 09 15:46:23 INFO: starting ip link set netns 3867304 docker0_0 name eth0 address 50:00:00:00:00:00 up
Cannot find device "docker0_0"
Mar 09 15:46:23 INFO: starting /opt/unetlab/wrappers/nsenter -t 3867304 -n ip addr add 172.17.136.141/16 dev eth0
Error: ipv4: Address already assigned.
Mar 09 15:46:23 INFO: starting /opt/unetlab/wrappers/nsenter -t 3867304 -n ip route add default via 172.17.0.1 dev eth0
RTNETLINK answers: File exists
PHP Warning: file_put_contents(/tmp/EVE | Login.desktop): Failed to open stream: Permission denied in /opt/unetlab/html/includes/cli.php on line 2078

[rusty725]

come to our chat and give me anydesk
https://webchat.eve-ng.net

[apekatt]

I had the same issue with installing EVE-NG Pro on our PowerEdge6715 running a AMD Epyc 9555P CPU. When selecting the bare metal option in grub it would just hang. There is a workaround that allows you to use the original eve-ng iso instead of using the Ubuntu legacy option:

1, boot the eve-ng pro iso
2, select "bare metal option" in grub, this will take you to a new page where you choose between "pro server" and "satelite"
3, press "e" to edit grub config

The original config (from grub.cfg) looks like this:

menuentry "Install EVE-NG Pro Server 6.5.0-21" {
set gfxpayload=keep
linux /casper/vmlinuz network-config=disabled quiet autoinstall ds=nocloud\;s=/cdrom/server/ ---
initrd /casper/initrd

Edit it to enable HWE:

menuentry "Install EVE-NG Pro Server 6.5.0-21" {
set gfxpayload=keep
linux /casper/hwe-vmlinuz network-config=disabled quiet autoinstall ds=nocloud\;s=/cdrom/server/ ---
initrd /casper/hwe-initrd

4, press F10 to boot

This worked for us, and seeing that we are both running PowerEdge-servers with 5th gen Epyc CPUs, it should work for you to. If you still have issues you could try a re-install with this workaround? No guarantees tho!

Info for reference: Dell PowerEdge 6715, AMD Epyc 9555P, 256GB RAM, 2TB HDD, BIOS firmware version 1.6.4. EVE-NG ISO: eve-pro-prod-bm-6.5.0-21-full

To the devs: Could this not be added to grub.cfg in the future? Ubuntu 22.04 is pretty old at this point and newer servers like the one me and OP are running will become more common. And the HWE option is allready present in the original eve-ng image. The reason i found this was basicly luck, i was desperately looking around in the .iso file and stumbled over the loopback.cfg in boot/grub where the HWE option was referenced, so i tried a hail mary and it actually worked!

[kazuki]

Thanks for the reply!
I actually ended up fixing it myself right after posting by doing the same thing.
I wasn’t seeing any issues with the basic functionality, but I did hit a problem where RDP wouldn’t work through Guacamole. (fixed 6.5.0-21)
Got the dev team involved, and it turned out the root cause was having too many CPU cores allocated, which triggered some unexpected behavior. So it wasn’t actually related to the HWE install.

Still, I’m pretty sure other people are gonna run into the same issue down the line, so thanks for writing it up and documenting it!

during troubleshooting they suggested switching to the EVE origin kernel, so you might want to try that too.
All you need to do is edit grub and set it to boot with the EVE origin kernel.

good lack.

[apekatt]

Everything seems to be working fine, but we are still in the process of installing various images and waiting for a license, so i haven't really tested it. The install went exactly as i should according to the cookbook, but we will see when we actually get labs and nodes up and running. We dont have a need for RDP, so if there is an issue there it wont be a problem.

The issue with the cpu-cores, that was cores allocated to nodes right?