openstack team mailing list archive

["Syd (Sydney) Logan" <slogan@xxxxxxxxxxxx>]

Hi,

Is there a recommended procedure for upgrading nodes that were configured pre-Folsom 3 to use quantum V1/OVS that were deployed with devstack? I probably should have asked this question before trying, but I went ahead and tried. I had a multi-node setup that I was driving with Horizon that was working very well. Now I'm just trying to get a single node setup working, and not getting far.

To get sync'd up with the latest, I did the following:

$ rm -rf /opt/stack (this is where devstack pulled things to)
$ rm -rf /etc/quantum; rm -rf /etc/nova

In the devstack localrc:

Removed n-net from ENABLED_SERVICES
Added q-dhcp to ENABLED_SERVICES (I had this disabled in pre-F3 after e-mails with Aaron Rosen when he helped me get going earlier, I've tried both ways and seems not to make a difference)
Added NOVA_USE_QUANTUM=v2 (but this doesn't seem to make a difference either)

And I ran devstack.

I got no errors when I ran devstack.

When I launched Horizon, some problems are evident. There is no launch instance button on the Instances page. Because I don't yet know the command UI enough to spin up and configure VMs, I figured I'd try running the devstack exercise.sh script to see what happens. It creates a few VMs, but none get an IP address (before I used to get IPs in 10.4.128.0).  It reports all tests passed, as well. If I click through in the UI on the VM, I see that for networking address it assigns all VMs is the value "Net1".

I've looked at console logs for the VMs created and see failures trying to dhcp (that's why I naively added q-dhcp back to ENABLED_SERVICES), but as I mentioned above, adding q-dhcp didn't help, and I'm wondering if it was a good idea anyway since Aaron steered me away from it before.

Output of ps shows expected services running (e.g., OVS daemon, plugins, agents) and services lists displayed by Horizon (e.g., nova, quantum, etc.) all seem normal to me.

Notably missing is the OVS gw- interface that was present before I upgraded (at http://wiki.openstack.org/RunningWQuantumV2Api there is this: "Note: with v2, Quantum no longer uses the L3 + NAT logic from nova-network. Quantum will not have the equivalent functionality until F-3, so you won't be able to ping the VMs from the nova controller host." Is that the reason?)  The gw interface is the way I could ping VMs from the host.

The missing gateway, horizon UI missing the create instance button, and not getting networks for VMs spun up by devstack's exercise script are the major symptoms.  I trust that devstack is up to sync with what is happening in Folsom, and that I am actually pulling down F3 code at this point (I've not tried to verify this).  I'm not aware of any need to tweek the devstack exercise script, I am assuming it is designed to work as is.

I'm thinking of wiping my entire disk and starting from scratch in case blowing away /etc/nova etc. and /opt/stack were not enough to reset state, but before I do this, any pointers to links or mail messages (I've scanned for relevant posts but missed finding any) that would be helpful before I do this?

Thanks,

syd