Openstack stop working after hosts reboot

Tue Mar 17 19:41:10 UTC 2015

I have one project with 10 instances. I’m ok with reinstalling everything, as far as I can save what’s inside Ceph (the volumes), because I have production data (logs) there.

Is that possible?

From: Andreas Hasenack [mailto:andreas at canonical.com] 
Sent: martes, 17 de marzo de 2015 16:38
To: Sacha Yunusic
Cc: ubuntu-openstack-installer at lists.ubuntu.com
Subject: Re: Openstack stop working after hosts reboot

On Tue, Mar 17, 2015 at 4:25 PM, Sacha Yunusic <sacha at penta-sec.com> wrote:

After some struggle (http://askubuntu.com/questions/596404/computers-not-reporting-to-landscape), Landscape is working fine and is managing hosts without problems.

In the other hand, on Openstack Dashboard I’m getting all kind of error messages like “Unable to delete volume snapshot: test snapshot”, “Can’t find floating IP”, etc.

Now I can start instances, but they are not reacheable at all.

My instances have an internal IP (part of the called “Admin_Net”, 10.10.0.0/16), and have associated a floating IP (part of the called “ext_net”, 10.222.221.0/24).

The Router inside Openstack that connects both networks is called “admin_router” and has IPs 10.10.0.1 (internal) and 10.222.221.2 (external).

Before I move my servers physically, I was able to reach everyithing. Now, I can’t.

If I launch an instance, and associate to that instance a floating IP, let’s say 10.222.221.10, I can’t reach it.

My guess was there is some wire problem, so I went to my Landscape environment’s bootstrap, called Balaguera (Physical server, IP 10.222.221.82 on the juju-br0 interface), I double check the cables on the NICs, both wires seemed to be OK, connected to the correct switch and everything.

I’m running out of ideas on what could be the problem.

PS: The good news is now I can attach and detach existin volumes to existing instances. I still can’t launch new instances. Actually, What I see when I try to launch a new instance, is the availability zone is “nova” instead of the one are all the others: “region1-1”.

Ps2: after writing rhis email, I remember to run the command you, Adam, told me to run:

stockachu at mistral:~$ JUJU_HOME=~/.cloud-install/ juju run --service landscape,landscape-msg 'sudo lsctl restart'

- Error: command timed out

  MachineId: 0/lxc/2

  Stderr: "Warning: Permanently added '10.222.221.140' (ECDSA) to the list of known

    hosts.\r\n"

  Stdout: ""

  UnitId: landscape/0

- MachineId: 0/lxc/3

  Stderr: "Warning: Permanently added '10.222.221.141' (ECDSA) to the list of known

    hosts.\r\nsudo: unable to resolve host juju-machine-0-lxc-3\n"

  Stdout: "== landscape-combo-loader ==\n... Not enabled, skipping.\n== landscape-appserver

    ==\n... Not enabled, skipping.\n== landscape-async-frontend ==\n... Not enabled,

    skipping.\n== landscape-job-handler ==\n... Not enabled, skipping.\n== landscape-msgserver

    ==\n * Checking database schema\n   ...done.\n * Restarting Landscape standalone

    message-server 1\n   ...done.\n * Restarting Landscape standalone message-server

    2\n   ...done.\n== landscape-pingserver ==\n... Not enabled, skipping.\n== landscape-api

    ==\n... Not enabled, skipping.\n== landscape-juju-sync ==\n... Not enabled, skipping.\n==

    landscape-package-upload ==\n... Not enabled, skipping.\n== landscape-package-search

    ==\n... Not enabled, skipping.\n"

  UnitId: landscape-msg/0

This would just restart the landscape services. What's odd is that the first one, "landscape", timed out.

	And now, If I go to “openstack” section in Landscape, it says “Installing region1 in nube_akainix” and some hook failed: "config-changed" errors (that’s on the GUI). The only option I see is to abort the installation, but I don’t want to do it. I don’t know if there is going to uninstall what was deployed before…

If you have a deployed cloud and something happened to the juju environment after that, the dashboard page in Landscape 14.10 would revert to the install page, as if the cloud was still being installed. This was fixed in LDS 15.01. If you hit "abort installation", it will most likely destroy the deployed cloud.

Can you deploy your cloud again using LDS 15.01? At this point, with so many changes and random errors, it might be the easiest solution. Or do you have many users and projects created there already?

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.ubuntu.com/archives/ubuntu-openstack-installer/attachments/20150317/6ad3df5d/attachment-0001.html>