auto_assigned_floating_ip - tools taking long time to bring back floating IP assigned

Bug #967166 reported by Kevin Jackson
22
This bug affects 4 people
Affects Status Importance Assigned to Milestone
OpenStack Compute (nova)
New
Undecided
Unassigned

Bug Description

ii nova-api 2012.1~rc1-0ubuntu2 OpenStack Compute - API frontend
ii nova-cert 2012.1~rc1-0ubuntu2 OpenStack Compute - certificate management
ii nova-common 2012.1~rc1-0ubuntu2 OpenStack Compute - common files
ii nova-compute 2012.1~rc1-0ubuntu2 OpenStack Compute - compute node
ii nova-compute-kvm 2012.1~rc1-0ubuntu2 OpenStack Compute - compute node (KVM)
ii nova-network 2012.1~rc1-0ubuntu2 OpenStack Compute - Network manager
ii nova-objectstore 2012.1~rc1-0ubuntu2 OpenStack Compute - object store
ii nova-scheduler 2012.1~rc1-0ubuntu2 OpenStack Compute - virtual machine scheduler
ii python-nova 2012.1~rc1-0ubuntu2 OpenStack Compute Python libraries
ii python-novaclient 2012.1~rc1-0ubuntu1 client library for OpenStack Compute API

in /etc/nova/nova.conf

--auto_assign_floating_ip

and I fire up an instance.
Everything works, a private IP is assigned... the instance is running... but it can take an inordinate amount of time (anywhere upwards of 2 mins, sometimes a lot longer ~10 mins) to associate a floating IP automatically.

On further investigation it appears the floating IP gets assigned almost immediately, the database table (nova.floating_ips) is updated - and I can use the floating IP to log in - but 'nova list' and 'Horizon' don't receive this information until a good few minutes wait.
The upshot is that instances are ready almost immediately, but to the end user they're not until they see the floating IP assigned.

Revision history for this message
Jason Kölker (jason-koelker) wrote :

The instance_info_cache needs to be updated after ip assignment. What's happening is the ip is assigned, but since all of compute/api use the instance_info_cache to get at network data, it won't show up until a get_instance_nw_info is called by compute (usually during the cache heal task).

Revision history for this message
Kevin Jackson (kevin-linuxservices) wrote :

Is there a way to force this, configure/change when this happens - is this code that needs to change?

Revision history for this message
Anthony Young (sleepsonthefloor) wrote :

This sounds like a duplicate of https://bugs.launchpad.net/nova/+bug/939122

Revision history for this message
Brano Zarnovican (zarnovican) wrote :

As Boris Deschenes pointed out on #939122, the patch did not fix the issue for auto-assigned floating IPs. There are possibly multiple places where the cache should be invalidated.

We are running Essex 2012.1.3 (fix for #939122 was included in 2012.1.1) and this issue is definitely there. Cache healing task will update only one instance at a time. If you have 20 instances running, it can take easily up to 50min until metadata server starts returning proper '/latest/meta-data/public-ipv4'.

To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.