creating ipv6 subnet on ipv6 vm will cause loss of connectivity

Bug #1616282 reported by Doug Wiegley
8
This bug affects 1 person
Affects Status Importance Assigned to Milestone
devstack
Fix Released
Critical
Brian Haley

Bug Description

The gate is all clogged, as any DSVM test using neutron (most of them), when it gets an IPv6 only OSIC node, seems to lose connectivity when devstack creates the internal neutron ipv6.

Not sure of root cause yet, but starting this bug here.

Please see starting at 01:48:

http://eavesdrop.openstack.org/irclogs/%23openstack-infra/%23openstack-infra.2016-08-24.log.html

Revision history for this message
Doug Wiegley (dougwig) wrote :
Revision history for this message
Armando Migliaccio (armando-migliaccio) wrote :

Do we know how to trigger/repro this, exactly?

Revision history for this message
Armando Migliaccio (armando-migliaccio) wrote :

fwiw I witnessed the resets on stable/mitaka changes too.

Revision history for this message
Armando Migliaccio (armando-migliaccio) wrote :

fwiw I witnessed the resets on stable/liberty changes as well.

Revision history for this message
Sean M. Collins (scollins) wrote :

The failure is related to RFC 6204. If we are launching nodes on an IPv6 cloud that uses RA adverts, and we are forwarding packets from the virtual infrastructure that exists on the node, we need to use the special sysctl flag accept_ra, and set it to 2. By default, Linux when ipv6 forwarding is enabled, will ignore RAs from other devices. Check out the RFC for details. The accept_ra = 2 is the workaround used for customer premises equipment which recieves RAs from upstream sources (like their ISP) and forwards packets.

Revision history for this message
Sean M. Collins (scollins) wrote :
Revision history for this message
Armando Migliaccio (armando-migliaccio) wrote :
Revision history for this message
Brian Haley (brian-haley) wrote :

Forgot to put the bug number in that review, but I'm not touching it now, will update if I re-spin.

Changed in neutron:
assignee: nobody → Brian Haley (brian-haley)
Revision history for this message
Doug Wiegley (dougwig) wrote :

359490 is now in the merge queue. Also not fixing the commit header.

Revision history for this message
Armando Migliaccio (armando-migliaccio) wrote :

I added the bug # in the stable backports commit message, if that's any help.

Revision history for this message
Armando Migliaccio (armando-migliaccio) wrote :
Changed in neutron:
status: Confirmed → Fix Released
affects: neutron → devstack
Revision history for this message
Richard Theis (rtheis) wrote :

This also impacted the networking-ovn native L3 job and was fixed by https://review.openstack.org/#/c/361420/.

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.