Bernd Bausch's profile - activity

You start by entering the error message into a search engine and select the results that best fit your problem.

Also, the sample programs I have seen include a call to set_missing_host_key_policy(). Perhaps that’s required?

I misread your question as “OpenStack SDK”, sorry. Paramiko is a library that implements SSH. If you can access the instance with SSH from the command line, it should also work with Paramiko.

The bridges connect containers to the network. Each container has its own set of interfaces, and they must somehow be connected to the physical network infrastructure that links the container hosts. This connection is done through bridges.

While I know very little about OSA, if I understand the documentation page correctly, you have to create them (brctl addbr) and attach them to a physical interface or a VLAN interface (brctl addif). See also https://docs.openstack.org/openstack-... (with graphics).

The SDK is for managing OpenStack objects - launch and delete instances, attach volumes, upload images etc. Access to instances is normally via network protocols such as ssh, http, etc, not via the SDK.

What is a vi instance? You are not referring to the text editor, right?

You can't live migrate from a failed node. You can however evacuate instances from that node and restart them.

There is an OpenStack project named Masakari whose goal is to provide instance HA. It is still at an early stage, as far as I understand it.

You don't know how fast the program would run on the physical server used by AWS.

What hypervisor is configured? On the compute node, check nova.conf. Most probably it's libvirt; in which case what is the value of the virt_type config parameter? If qemu, your cloud emulates CPUs, thus slow.

This is confusing; I don’t understand what the placement user is doing here. What are your OS_... environment variables set to?

Since this is a different problem, I suggest moving it to a new question.

I don’t have an answer, but it seems you can still go back to the old way https://askubuntu.com/questions/10317.... Or here: https://netplan.io/faq#how-to-go-back....

There seems to be a migration tool, too: netplan ifupdown-migrate.

I can’t see your answer file because ufile.io doesn’t work with mobile Apple devices, but did you follow these instructions: https://www.rdoproject.org/networking...

I.e. you need to insert your home network interface in br-ex.

As long as they are on the same layer 2 network, i.e. can talk to each other with MAC addresses, it should be OK. Everybody uses virtual machines connected with virtual switches to deploy OpenStack for fun and study.

DHCP servers don’t need to run on compute nodes but can. Linuxbridge agents must.

Something wrong with the connection between compute and controllers. Share computes’ ML2 and agent config (e.g. paste.openstack.org).

You can run OpenStack on a single node without switch, but I doubt you want this answer. I think that OpenStack requires nodes to be on the same network, but if they are not connected via switch, perhaps you could connect them to an overlay network. How are your nodes connected if not via switch?

Floating IPs are managed by routers connected to provider networks, so the second answer is no. See also https://assafmuller.com/2018/07/23/te....

Yes I should have said Linuxbridge configuration instead of ML2. The installation guide doesn't use Openvswitch nor SRIOV though. Could the problem be there?

When you get a 500 error listing Neutron agents something more fundamental than DHCP is wrong. Check the Neutron logs for clues.

Please use the 101010 button to turn your code into something readable.

The IBM instructions are for Juno, 4 years old or so. The driver config setting looks incorrect. Try setting it to ldap. See https://docs.openstack.org/keystone/l....

Instances get IP addresses via DHCP. By default, one DHCP server runs per network. Either the DHCP server is down or a network problem blocks DHCP.

openstack network agent list shows the DHCP server status.

Set the address manually and test connectivity to the router. Double-check ML2 config.

Not that easy. You can, in principle, run a heterogeneous cloud consisting of Rocky Nova, Queens Cinder etc. This is even supported during an upgrade: You upgrade one physical node after the other, while the cloud is running, and during the process you have a mix of version x and version x+1.

But of course, you don't want to mix versions in a production cloud, except for the short upgrade window.

You use nova-manage, cinder-manage etc to get the versions of each service, but these commands access the database and print the database version rather than making an API.

The best bet IMO is to retrieve the API version. Those services that support microversions will also provide the most recent microversion, from which you can derive the OpenStack version. For example Nova:

$ curl http://192.168.1.200/compute | python -m json.tool
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100   381  100   381    0     0  27846      0 --:--:-- --:--:-- --:--:-- 29307
{
    "versions": [
        {
            "id": "v2.0",
            "links": [
                {
                    "href": "http://192.168.1.200/compute/v2/",
                    "rel": "self"
                }
            ],
            "min_version": "",
            "status": "SUPPORTED",
            "updated": "2011-01-21T11:33:21Z",
            "version": ""
        },
        {
            "id": "v2.1",
            "links": [
                {
                    "href": "http://192.168.1.200/compute/v2.1/",
                    "rel": "self"
                }
            ],
            "min_version": "2.1",
            "status": "CURRENT",
            "updated": "2013-07-23T11:33:21Z",
            "version": "2.65"
        }
    ]
}

The very last line, "version": "2.65", shows that this is Rocky.

What’s the file size? Do you get a timeout when you use the openstack client?

It’s not quite clear what times out. Try separating image creation (“reservation”) and upload.

Third, did you run stack.sh on a freshly installed system? If your system contains any Python libraries that are out of date or otherwise incompatible, anything can happen. Start from scratch: Reinstall the OS, clone the repo and run stack.sh.

Next, is there another error or warning message earlier in the log? Of course, given your local.conf, you have no log file. Add log file directives to local.conf and run stack.sh again. Look for other errors or warnings.

I have several thoughts. First, a glitch? You are installing from the master branch of devstack, and stack.sh copies the master branch from Keystone. Perhaps it's unstable at times. Have you considered using the stable/rocky branch of devstack? Add option -b stable/rocky to the clone command.

I think it actually does, at least for Rocky. See the git repository.

It might be that the method you use to install Ceilometer configures a different pipeline.

Try formatting the output as code using the "101010" button, otherwise it's just a wall of unreadable text.

What is your local.conf? What command did you use to clone Devstack? Are you installing it on a freshly installed system? What distro do you use to run Devstack?

There is no service libvirt-bin. There is, however, a libvirt-bin software package. If that's what you mean: nova-compute manages VMs, and it uses libvirt to do that in most cases.

You can configure Nova so that it uses a different VM management method, for example native Xen, but most OpenStack clouds configure libvirt.

I would suspect some other device has the same IP address.

setup.cfg and setup.py are part of the software setup framework of all Python packages, not only in OpenStack.

Zuul is a central part of OpenStack testing and CI.

The project has a documentation site. You can also use the openstack-discuss mailing list; include [ansible-hardening] in the subject. They should have an IRC channel, too, but I currently don’t see it documented.

The point of a volume is that it conserves the data on it. If you want to wipe it, you have to do that yourself, but again, that defeat’s the volume’s purpose.

Why don’t you want to delete the volume?

What exactly do you do to get this exception? Are you running programs that use OpenStack’s Java SDK?

Looks like ceph.conf is not found. It doesn’t exist or is not at the expected location.

For all instances that you launch with mykey. But if you like you can create other keypairs and run instances with those.

110 means "connection timed out". I guess Ceph is not running or not reachable.

Best practice is to not have passwords on cloud instances. If you have several hundred instances to look after, you either give them the same password (bad for security) or you have to remember several hundred passwords (good luck).

So, the standard practice is to use password-less ssh. You set up a keypair (openstack keypair create mykey > mykey.pem), launch the instance with the key (openstack server create --key-name mykey ...), and ssh into the instance using the private key (ssh -i mykey.pem ...).

i suggest you learn about ssh and keys, and Nova and keypairs.

I have deployed Docker on VMs with one CPU and 2GB memory. Docker doesn’t load your system. What loads your system is the applications you run in containers. You know best what their requirements are.

There is an issue with your network config or metadata server. Does the instance have any network connectivity? Is the metadata service running? Check with openstack network agent list.

Not sure what the ubuntu module is about. I will have a look at your parameters; don’t hold your breath.

I use it and it works. On your instance, can you curl http://169.254.169.254? What is the precise error message from cloud-init? What user data do you send to the instance?

You say:

systems/applications inside openstack instances using docker container

If you plan to run instances as Docker containers, you need to use the Nova Docker driver. I don't think it is still supported, though. Linux container instances (using LXD) are supported and are promoted by Canonical.

If you want to run Docker on instances, just deploy Docker on the instances. How? Any way you like; instances are just like physical machines as far as Docker is concerned. There are many options, but you can start at the Docker quick start series of web pages. Or use Magnum to deploy a Docker Swarm, Kubernetes or Mesos cluster on instances.