AWSVPC mode invalid route table when scaling

[panga]

I'm having issues of bad route table when I add/remove new nodes (EC2 instances with Weave in awsvpc mode) with Auto Scaling Groups.

My auto-scaling group has minimum 1 and maximum 3 instances.
After some auto scaling I have 2 instances running, but only 1 valid route in the table and a lot of Black Holes.

Destination	Target	Status
10.32.0.0/13	eni-57f0ae51	Black Hole
10.40.0.0/13	eni-91db8297 / i-c4c017d2	Active
10.40.0.0/14	eni-54f0ae52	Black Hole
10.44.0.0/15	eni-54f0ae52	Black Hole
10.46.0.0/15	eni-6d673e6b	Black Hole

BTW I tested many times and always end with invalid route tables.

[panga]

Steps to reproduce:

• Start a EC2 instance of Weave AMI ami-f66607e1.
• Configure weave router to use AWS VPC and create a new image:

$ sudo yum install -y vim
$ sudo vim /etc/weave/run.sh
change line: succeed_or_die weave launch-router --awsvpc $PEERS
$ sudo stop weave
$ weave reset --force
$ sudo start weave

• Create an auto scaling group with previous created image and add/remove nodes simulating elastic scaling.
• Look at the AWS VPC route tables and see the invalid configuration.