Problems running flocker acceptance tests

[sheeta...@gmail.com]

Hi,

We have a flocker storage plugin which passes all the trial tests. After doing that we were able to setup two nodes with the appropriate .crt and .key. This was done with the manual flocker install. The flocker nodes can run the following commands -

docker run -v apples:/data --volume-driver flocker busybox sh -c "echo hello > /data/file.txt"

docker run -v apples:/data --volume-driver flocker busybox sh -c "cat /data/file.txt"

The volumes get created on our storage backend and flocker is able to move the volume from one node to the other.

The next step for us is to run the acceptance tests. I am not sure if we need to setup the flocker dev environment on the flocker nodes or a different machine. For now I have it setup on a non flocker machine. The config.yml looks like this -

metadata:

  creator: Me

osnexus_flocker_driver:

  username: "admin"

  password: "password"

managed:

  addresses:

    - "10.0.11.41"

    - "10.0.11.42"

storage-drivers:

  osnexus_flocker_driver:

    backend: "osnexus_flocker_driver"

    username: "user"

    password: "password"

admin/run-acceptance-tests --keep --distribution=ubuntu-14.04 --provider=managed  --config-file=/root/flocker/config.yml --flocker-version=1.10.2 --dataset-backend=osnexus_flocker_driver

admin/run-acceptance-tests --keep --distribution=ubuntu-14.04 --provider=managed  --config-file=/root/flocker/config.yml --flocker-version=1.10.2

The above commands change /etc/flocker/agent.yml on both the flocker nodes -

cat /etc/flocker/agent.yml

control-service: {hostname: 10.0.11.41, port: 4524}

dataset: {backend: zfs}

version: 1

I am not sure whats the right way to run the acceptance tests.

Thank you in advance !

--Sheetal

[DiStephane]

Hi Sheetal,

I tried to do data migration with flocker using ZFS backend. But I have not get the desired result. Since you have managed to achieve this step can you help me?

[Madhuri Yechuri]

Hi Sheetal,

> admin/run-acceptance-tests --keep --distribution=ubuntu-14.04 --provider=managed  --config-file=/root/flocker/config.yml --flocker-version=1.10.2

Running above command without "dataset-backend" option set default to ZFS: https://github.com/ClusterHQ/flocker/blob/master/admin/acceptance.py#L936

> admin/run-acceptance-tests --keep --distribution=ubuntu-14.04 --provider=managed  --config-file=/root/flocker/config.yml --flocker-version=1.10.2 --dataset-backend=osnexus_flocker_driver

Can you please reset your agent.yml on the Flocker Agent nodes and retry with "dataset-backend" option set to your backend?

Thanks!

madhuri.

[Ryan Wallner]

DiStephane,

This thread is using a new backend sheetal is making "osnexus_flocker_drive", she is not using ZFS, please keep the conversation seperate from this issue thanks.

[sheeta...@gmail.com]

Hi,

Please find the attached output of the acceptance tests executed on the flocker control service node. Running the tests reboots the node and the tests are never executed. I setup the flocker dev environment on both the flocker nodes. The same thing happens if I run the tests on the other flocker node. Do I need to setup the acceptance tests on clean boxes without prior flocker installation ?

Thanks,

Sheetal

[sheeta...@gmail.com]

Thank you Madhuri and Ryan for your responses. I am now able to run most of the acceptance tests. I will post specific questions on the tests which fail.

Thanks,

Sheetal

On Thursday, March 31, 2016 at 1:29:00 PM UTC-7, sheeta...@gmail.com wrote:

[Leibo]

Hi,

Can you please share what was eventually setup in order to run the acceptance tests?

What needs to be installed on the managed nodes and the test execution server? where do I run the tests from (flocker nodes? third host?)

Any relevant information would help, thanks

[Leibo]

Just to add...

My setup includes one server with the Flocker dev env and 2 servers for the acceptance tests (with no Flocker installation)

I have password free SSH access from the dev server to both nodes.

When trying to execute the tests It hangs on "Saving dataset backend config to: /tmp/tmpgCW7mg/dataset-backend.yml"

In the tests run log I see:

Disconnecting with error, code 14\nreason: no more authentication methods available", "message_type": "twisted:log", "task_level": [1]}

[Richard Wall]

On Sunday, November 27, 2016 at 8:31:43 PM UTC+2, Leibo wrote:
> Can you please share what was eventually setup in order to run the acceptance tests?
> What needs to be installed on the managed nodes and the test execution server?
> where do I run the tests from (flocker nodes? third host?)
> Any relevant information would help, thanks

There is documentation explaining how to run acceptance tests on "managed" nodes.
 * https://flocker-docs.clusterhq.com/en/latest/gettinginvolved/acceptance-testing.html#managed

You'll need to have two managed nodes.
All the managed nodes need to have Docker, flocker-dataset-agent, flocker-docker-pluging and your storage plugin pre-installed and running.
One of the managed nodes needs to have the flocker-control running.
Run `run-acceptance-tests --managed` from your laptop or from any computer that can connect to the managed nodes.
It will create new Flocker API certificates and reconfigure Flocker (and Docker) each time you run the tool.
It needs to be able to SSH into the nodes as root in order to upload the certificates.
So you must have enabled SSH key based log in for the root user and you must make sure that there is an SSH agent with the private SSH key loaded before running `run-acceptance-tests`.

On Monday, 28 November 2016 16:52:08 UTC, Leibo wrote:
> Just to add...
> My setup includes one server with the Flocker dev env and 2 servers for the acceptance tests (with no Flocker installation)
> I have password free SSH access from the dev server to both nodes.
> When trying to execute the tests It hangs on "Saving dataset backend config to: /tmp/tmpgCW7mg/dataset-backend.yml"
> In the tests run log I see:
> Disconnecting with error, code 14\nreason: no more authentication methods available", "message_type": "twisted:log", "task_level": [1]}

That sounds like an SSH authentication problem.
I've only ever used `run-acceptance-tests` from Linux.
If you're on windows, perhaps `run-acceptance-tests` script can't communicate with your SSH agent.
There's also some discussion here from another user who encountered that error:
 * https://github.com/ClusterHQ/flocker/issues/2874

Unfortunately we never got any more diagnostic feedback from them.

-RichardW.