Rancher: Error with pre-create check: "unexpected end of JSON input"

[jmondragon]

Rancher: Error with pre-create check: "unexpected end of JSON input"

[lnxbil]

Hi,

can you try the latest prerelease version? It has improved error reporting and was tested with Rancher OS extensively and may indicate what the problem is.

[jmondragon]

I did end up trying the latest, and it gave more information. It gave me the following error:

Flag provided but not defined: -proxmoxve-storage-type;

Seemingly randomly choosing the flag (e.g. -proxmoxve-user, -proxmoxve-storage, etc.)

I went back to v2 and changed my HOST from the IP address to a hostname, and it seemed to work better. I wonder if it has something to do with the . in IP addresses or FQDN.

I still ended up having trouble connecting via SSH, but the VM did get created. I will continue to troubleshoot.

Thanks,
Josh

[CarlosLVar]

Hello there,

We too have encountered the same problem when using v2, will try to provide a hostname instead of an IP as well.

Also, we are experiencing the other problem with v3, but I recon that should probably go in another issue, though logs indicate just that the mentioned flag is empty whereas others aren't (and the host doesn't change if I put an IP, it takes the default 192... Would have to check with a hostname as well.

Thank you for your work!

[ropeguru]

Testing the v3pre3 and every run of trying provision a node gives me a different error like:

Flag provided but not defined: -proxmoxve-disksize-gb; Timeout waiting for ssh key

Each run, or even each node trying to be deployed, will give an error like this. Sometimes it is -proxmoxve-disksize-gb or -proxmoxve-guest-ssh-port or -proxmoxve-storage. Seems to just pick a random defined entry to fail on.

Really looking forward to getting this working.

Edit: I originally tried the v2 and also had issues. Running Rancher v2.3.3 and Proxmox 6.1-5

[ropeguru]

Adding some additional info..

It appears that when Rancher reads the template to gather defined values, something is going wrong.

For instance, I have "NFS-Datastore" defined for the storage location. In the log below, it is showing as "local". Also I have the disk size configured for 50GB, but the entry below is 16Gb. And the list goes on.

Debugguing Rancher container, I find the following in the log: (passwords have been changed)

2020/01/07 18:35:04 [DEBUG] create cmd [create -d proxmoxve --engine-install-url https://releases.rancher.com/install-docker/19.03.sh --proxmoxve-user root --proxmoxve-driver-debug --proxmoxve-image-file NFS-Datastore:iso/rancheros.iso --proxmoxve-password password --proxmoxve-disksize-gb 50 --proxmoxve-guest-ssh-port 22 --proxmoxve-memory-gb 8 --proxmoxve-storage-type qcow2 --proxmoxve-guest-username docker --proxmoxve-host 192.168.1.171 --proxmoxve-realm pam --proxmoxve-storage NFS-Datastore --proxmoxve-guest-password 123456]
2020/01/07 18:35:04 [INFO] Provisioning node ranchnode1
2020/01/07 18:35:04 [DEBUG] stdout: Incorrect Usage.
2020/01/07 18:35:04 [INFO] [node-controller-docker-machine] Incorrect Usage.

So something in the create cmd under Rancher is not correct. The values passed in the first entry are what I have in my template.

Edited for more constructive info.

Another edit: I found the issue by manually running the create command in my Rancher container.

It seems that the v3-pre3 is still using the option form of --proxmox- but the options being generated in a Rancher template are in the form --proxmoxve-. That is why it is failing.

If I manually run a machine-create in the Rancher container using the v3-pre3 driver, I still get the json error, but it does connect and pull the next valid ID.

Hope this helps.

[lnxbil]

New version with a lot of fixes and merged PR. Please try again and report back.

[jmondragon]

Thank you, I'll give it a try. I don't see the docker-machine-driver-proxmoxve.linux-amd64 (on the v3 release page), only zip and tar.gz files. Can I use the link to the tar.gz within Rancher?

[lnxbil]

Thank you, I'll give it a try. I don't see the docker-machine-driver-proxmoxve.linux-amd64 (on the v3 release page), only zip and tar.gz files. Can I use the link to the tar.gz within Rancher?

Ah sorry, I clicked the "published" button too soon. The binaries weren't completely uploaded yet. I reuploaded them and now they should be present.

[jmondragon]

This worked great! I did have to use the latest 1.5.5 rancheros-proxmoxve-autoformat.iso as outlined in the README.

[cyrus104]

I am using V3 and am getting this error.

I currently have a RancherOS running in one VM on a 3-node proxmox cluster, inside RancherOS docker has running Rancher that has this driver load. When I try to create a new cluster with a node, I get this same exact issue.

[mjkl-gh]

I'm running into this issue as well on V3 using both the latest (1.5.6) and 1.5.5 image. I have no idea where to begin debugging this or which information to provide.

I've tried this from a windows 10 laptop (inside ubuntu wsl1 20.04) and regular ubuntu 20.04 both with the same result (on the same pve node)

edit:

Just noticed that go-resty is returning a "596 tls_process_server_certificate: certificate verify failed" after which proxmoxve driver is returning the unexpected end of JSON input"

This is weird since I set up my CA correctly for my hostname (using Let's encrypt). It returns the same error when using the IP address instead of hostname.

[lnxbil]

Can you please test with v4? I already could not replicate the issue with v3.

[mjkl-gh]

Aiaiai. I must unfortunately admit I remember fixing this issue. If I'm working from memory it was something stupid as using the wrong pve node or misspelling a hostname. I will try to get back to you on this!

[cedvan]

Same problem with v4.

I have test deploy with docker-machine manually with same parameters, and works!
But from rancher UI failed with Timeout waiting for ssh key and no other information to help debug 😢

[cedvan]

Hum ok, problem with dns resolution, use IP in host and works. It's sad but work...

Be careful, during my tests, the creation of VMs failed but yet I had residual vm-disks, be sure to check and delete them manually if necessary. Otherwise you will have 500 during the next creations because vm-disk already exist