When checking CAPI Trail for the CreateNode action and the error_detail you note a stack trace referencing "AzureComputeVirtualMachine" on line 336 as well as references to the java class "SecureShellHelper".
Release or Environment
In ecc_agent_script_include AzureComputeVirtualMachine we try to set the loopback address for Linux VMs. We first ping the VM to see if it's available and then initiate a SSH session to do so. Occasionally though the VM can be reachable by ping but still unavailable on its ports. So we're unable to make a connection and set the loopback.
Because this fails, the entire Stack provisioning fails even though the VM is properly deployed.
A crude remediation is to add a sleep so that we wait until the VM is fully available.
- Navigate to AzureComputeVirtualMachine in ecc_agent_script_include
- Go to line 333
- Create a new line above (first line after start of else block) (before var hostNameLoopBackUpdateScript)
- Add following code:
You can adjust the sleep as long or short as necessary