Bernd Bausch's profile - activity

I don't know whether Kolla supports them, but generally they are "supported" in the sense of "tested" and "released to production". Find documentation under docs.openstack.org/magnum, docs.openstack.org/sahara and docs.openstack.org/murano.

You want to upload a file to a different account? I don't know if this works. Probably not, as you have to provide an authentication token that matches the account.

I suggest you create a new question.

I don't know much but assume it will solve itself once the replicas are recreated.

Network isolation is achieved with the ML2 type driver (assuming that ML2 is the core plugin). Either VLANs or one of the tunnel technologies GRE, VXLAN or Geneve.

DHCP namespaces don't separate tenant networks. Their purpose is to have several network interfaces with identical IP addresses on the same controller.

1- I think so.

2- I would not bother separating three nodes. Three hundred, yes.

3- I guess it only reports that replicas are missing. I'd expect this to happen when the disk is offline.

Yes you can have servers dedicated for objects and other servers dedicated for accounts and containers. I suppose you would run the ring builder commands with different sets of IPs.

2- I think you have to write your own authentication solution. Perhaps https://docs.openstack.org/swift/late... helps.

4- https://stackoverflow.com/questions/1...

1. My guess is that it is not defined what happens when you specify both

2. I don't know, but most likely not in a database but inside some process' memory

I am not 100% sure if I understand all details correctly, but I am confident that my explanation is correct at a high level.

Keystone middleware is a Python program used by other components of OpenStack to perform authentication. It validates tokens.

You need to understand how authentication works. E.g. when you log on to the GUI, it obtains a token for you. At each action, e.g. launching an instance, listing volumes, creating a router, the same token is submitted to Nova, Cinder and Neutron. It needs to be validated by talking to Keystone.

Rather than validating the same token again and again, the keystone middleware caches it, normally in memcached. Swift has its own token cache, and the swift.cache option tells the keystone middleware to use the Swift cache instead of memcached.

If you use memcached as the caching method, Swift's token caching is done separately from the caching of other OpenStack services.

To go back to the original question, you set these options in the hope that sharing the token cache with Swift will be beneficial for performance.

Read this first: https://docs.openstack.org/keystone/p....

I don't think "private IP" is a Neutron term. "static IP" is better. It's an address from a Neutron network that was created by a user, a.k.a "tenant network". "external network" is a Neutron network that corresponds to a network outside the cloud. Floating IPs are obtained from external networks.

It doesn't work because there is no route to 172.16.10.4. I don't know why the VM's floating IP doesn't come out of the floating range, but I would remove the FLOATING_RANGE option and set up a new stack. Ideally on a fresh OS installation.

Your public-subnet is 172.24.10.0/24, and the floating IP is 172.16.10.4? How is that possible?

You ping from where?

Note that 172.16.0.0/16 is a private address range that doesn't work on the internet. By default, Devstack uses it for a fake external network.

I am grateful for the opportunity to learn. For one thing, I had thought that no other open-source clouds remained besides OpenStack. Good to see that this is not the case.

Step 4 is rather significant. Thanks for pointing it out.

I suppose you can't associate a floating IP because the instance is connected to the public network. Remember that floating IPs are taken from public-subnet - and your instance has such an address already!

It seems to me that you launched your instance on the public network, which most probably doesn't have a dhcp server. Thus, the instance doesn't get an IP address.

You can check for dhcp with openstack network show public. You can also check the instance's console log if it received an address.

Can you share the precise command for launching your instance and the precise error message?

It's probably not enabled by default. Check if python-networking-sfc is installed; if not, install it and try it.

Assuming the instance uses ephemeral storage, and that OpenNebula knows what to do with qcow2, I would try this:

1. take a snapshot from the instance
2. download the snapshot image to a qcow2 file
3. use that file in the OpenNebula cloud to launch an instance

It just means that instances are implemented by virtual machines in a vsphere cluster. Why would you want that? Perhaps you have a vsphere installation and want to integrate it with OpenStack services such as Magnum (managing Kubernetes clusters), Sahara (Big Data), Trove (database servers) etc. Or you plan to write applications using the OpenStack API. Or you want to gradually move from Vmware to OpenStack. A few use cases by Vmware itself: https://www.vmware.com/products/opens....

What did you do when the error occurred?

Perhaps it's configured to log to stderr, in which case you will probably find the logs in the systemd journal.

Config guide: https://docs.openstack.org/placement/....

No. This file contains messages generated by Apache. Where your Placement log is depends on your configuration; probably somewhere under /var/log.

What do you mean by "old image in OpenStack"? There is no Cirros image in OpenStack. All images are on https://download.cirros-cloud.net/.

There is no screenshot.

I suppose the line number refers to one of the shell scripts that are launched by stack.sh, perhaps stack.sh itself. The script name should be mentioned.

Answered at https://ask.openstack.org/en/question....

The error message on the command line means that you didn't tell the openstack client how to contact Keystone. There are two ways to do that: With a command line option --os-auth-url, or an environment variable OS_AUTH_URL. See https://docs.openstack.org/python-ope....

On your controller, the environment variable can be set by sourcing admin-openrc or demo-openrc.

Furthermore, the error in the Apache log file indicates that your web server is unable to contact Glance. I guess the Glance API process is not running at that very moment.

I also think that the two messages are unrelated. The error log contains a request to list images, not to create an image. Does the timestamp in the Apache log correlate with the image create command above?

May I delete logical volume or extend it?

If you delete cinder-volumes-pool, Cinder won't work. You can extend it, but there is no need, as nothing has been allocated from the pool so far. It has almost 20GB free space.

I created a cinder volume without loop device on /dev/sdb1 below

No. You added an LVM (not Cinder) physical volume to the cinder-volumes volume group. What happens when you try openstack volume create --size 1 newvol? Does it work? If not, any errors in Cinder logs?

privsep-helper can't be found. It is either not installed or not in the execution PATH.

I suggest you edit your question and format the log messages so that they become readable. Use the button labeled "101010".

Seems the Placement service is not running. Check the placement logs for reasons why.

I don't know this either. Use the documentation web site and/or the help feature of the openstack client.

See https://docs.openstack.org/image-guid....

There might be a problem with the loop device, yes, but I can't see anything wrong in the pvs/vgs/lvs output. Again, find out which command failed.

What does losetup say?

Can you create Cinder volumes manually?

Can you create logical volumes in the cinder-volumes from cinder-volumes-pool?

I doubt that launching an instance without volume leads to an error message that the volume can't be created. In other words, the BUILDAbortException you found in the Nova log is not related to the instance without volume.

If you don't have the log anymore, it's hard to troubleshoot a past problem. I would provoke the error again, this time taking care to save the log messages.