Hi,
Thanks for looking into that. Indeed, while trying to reproduce the
issue this morning, I found it more challenging than I originally
thought. I want to add a few points here on how I reproduced the issue:
1. Usually, I do not use the Azure CLI directly. I use a custom CLI of my own
that uses the Azure SDK. This custom CLI always creates the NIC (with AN)
before creating the VM. The VM is created with the existing NIC. I don't know
how Azure CLI manages "--accelerated-networking" flag under the hood, maybe
it's doing something different that makes it harder to reproduce the issue.
2. (for costs reasons) I always create a new resource group when I create a
new VM. Once again, I don't know if it has any impact on the reproducibility.
3. This morning I managed to reproduce the issue using only the Azure CLI and
after a few (unsuccessful) tries:
➜ ~ az group create --resource-group hirsute-acc-manual-1 --location 'UK South'
{
"id":
"/subscriptions/5059ce5a-a72d-4085-acb7-33b421daa1ee/resourceGroups/hirsute-acc-manual-1",
"location": "uksouth",
"managedBy": null,
"name": "hirsute-acc-manual-1",
"properties": {
"provisioningState": "Succeeded"
},
"tags": null,
"type": "Microsoft.Resources/resourceGroups"
}
➜ ~ az vm create --name hirsute-acc-manual --resource-group
hirsute-acc-manual-1 --location "UK South" --image
'Canonical:0001-com-ubuntu-server-hirsute-daily:21_04-daily-gen2:latest' --size
Standard_F8s_v2 --admin-username ubuntu --ssh-key-value "$(cat
~/.ssh/canonical.pub)" --accelerated-networking
{- Finished ..
"fqdns": "",
"id":
"/subscriptions/5059ce5a-a72d-4085-acb7-33b421daa1ee/resourceGroups/hirsute-acc-manual-1/providers/Microsoft.Compute/virtualMachines/hirsute-acc-manual",
"location": "uksouth",
"macAddress": "00-22-48-40-82-32",
"powerState": "VM running",
"privateIpAddress": "10.0.0.4",
"publicIpAddress": "51.104.198.218",
"resourceGroup": "hirsute-acc-manual-1",
"zones": ""
}
➜ ~ ssh -i ~/.ssh/canonical ubuntu@51.104.198.218
The authenticity of host '51.104.198.218 (51.104.198.218)' can't be established.
ECDSA key fingerprint is SHA256:wIQAUjmIeFvdBeqT5a2RHJEpDtjCnrJ+FggR8pzW7OM.
Are you sure you want to continue connecting (yes/no/[fingerprint])? yes
Warning: Permanently added '51.104.198.218' (ECDSA) to the list of known hosts.
ubuntu@51.104.198.218: Permission denied (publickey).
4. I will post the full syslog file here but I also want to point that
I THINK this issue only appears with mlx5 devices/drivers. When I was
checking the VM created with no issue, mlx4 modules were loaded. On the
previous VM, I can see:
ubuntu@hirsute-acc-manual:~$ lsmod | grep mlx
mlx5_ib 331776 0
ib_uverbs 139264 1 mlx5_ib
ib_core 348160 2 ib_uverbs,mlx5_ib
mlx5_core 1081344 1 mlx5_ib
tls 90112 1 mlx5_core
mlxfw 36864 1 mlx5_core
Once again, I don't know if that really matters.
** Attachment added: "syslog"
https://bugs.launchpad.net/cloud-init/+bug/1919177/+attachment/5487677/+files/syslog
--
You received this bug notification because you are a member of Kernel
Packages, which is subscribed to linux-azure in Ubuntu.
https://bugs.launchpad.net/bugs/1919177
Title:
Azure: issues with accelerated networking on Hirsute
Status in cloud-init:
Incomplete
Status in cloud-init package in Ubuntu:
New
Status in linux-azure package in Ubuntu:
New
Bug description:
[General]
On Azure, when provisioning a Hirsute VM with Accelerated Networking
enabled, sometimes the SSH key is not setup properly and the user
cannot log into the VM.
[how to reproduce]
Start a VM with AN enabled:
```
az vm create --name "$VM_NAME --resource-group "$GROUP" --location "UK South"
--image
'Canonical:0001-com-ubuntu-server-hirsute-daily:21_04-daily-gen2:latest' --size
Standard_F8s_v2 --admin-username ubuntu --ssh-key-value "$SSH_KEY"
--accelerated-networking
```
After a moment, try to SSH: if you succeed, delete and recreate a new
VM.
[troubleshooting]
To be able to connect into the VM to debug, run:
```
az vm run-command invoke -g "$GROUP" -n "$VM_NAME" --command-id
RunShellScript --scripts "sudo -u ubuntu ssh-import-id $LP_USERNAME"
```
In "/run/cloud-init/instance-data.json", I can see:
```
"publicKeys": [
{
"keyData": "<my-pub-key>",
"path": "/home/ubuntu/.ssh/authorized_keys"
}
],
```
as expected.
To manage notifications about this bug go to:
https://bugs.launchpad.net/cloud-init/+bug/1919177/+subscriptions
--
Mailing list: https://launchpad.net/~kernel-packages
Post to : kernel-packages@lists.launchpad.net
Unsubscribe : https://launchpad.net/~kernel-packages
More help : https://help.launchpad.net/ListHelp
