Bonding Debian Juniper

[Volker Lieder]

Hi Support,
we try to enable bonding with debian 9.3 and proxmox Virtual Environment 5.1-46.
The bonding is about 2 x 10Gbit interfaces and we try to use mode 802.3ad.
The lacp is active and stable as long as we shutdown one link on juniper switch for testing.
When the port wents down, we recieve kernel errors and the system starts to freeze.
If we enable the port on juniper again, the lacp wont get in stable modus.
First if we reboot the server, lacp is up and running again.

On cli we see the following messages:
kernel:[ 5624.078761] watchdog: BUG: soft lockup - CPU#9 stuck for 22s! [kworker/u64:0:8324]

Kernel messsages attached in a .pdf file.

The system is new installed and we do some failover tests before put it in production mode.

Regards,
Volker

 

[wolfgang]

Hi
This looks like an driver problem of the i30e module.
You can try the new intel driver from the Intel download-center

https://downloadcenter.intel.com/do...bit-Ethernet-Network-Connections-Under-Linux-

or test the new PVE 4.15 kernel.

 

[Volker Lieder]

Hi, thank you, new driver was no success, on 1Gbit Bond on the same machine everything is fine.
I will test PVE 4.15 come back to this thread with results.

 

[wolfgang]

Some Intel networkcards can update firmware.
So check also if it is possible for this NIC.

 

[Volker Lieder]

It seem it is a supermicro card with an intel chipset. All available drivers and firmwares are up2date.
I am awaiting subscription for this project to test new kernel.

 

[Volker Lieder]

Hi, we have subscription for this setup now, how can we donwload/install Kernel 4.15?
I dont see it in Repo.
Regards,
Volker

 

[Volker Lieder]

i got it. wrong repo path. With Kernel 4.15 the same errors occurs. I think we have to change bonding type for the moment. Or any other ideas?

 

[wolfgang]

This kernel is in the testing repo

see
https://forum.proxmox.com/threads/4-15-based-test-kernel-for-pve-5-x-available.42097/

 

[wolfgang]

Or any other ideas?

I have no more ideas at the moment.

 

[Volker Lieder]

We tryed on another server with dual intel cards, here everything works as expected. We are in contact with supplier, perhaps we can get another firmware version from supermicro. At least we can change to other network cards. If we get success with this card i will report it here
thank you

 

[Volker Lieder]

Hi, we tried new firmware from supermicro with no success.
Only way we can use with Supermicro AOC-STG-i4S is Active-Passive.
All other modes end up in a kernel bug. Also with another "normal" debian kernel, so its not a proxmox issue.
Just for knowledge base, from our site Supermicro AOC-STG-i4S is not usable in a proxmox lacp/bonding environment at the moment.
We think about use active-passive or change network cards.
Thank you for your support.

 

[wolfgang]

Thanks for the feedback.