GCP, basic IPv6 set up for Linux VM Instance

I'm trying to set up a simple VM in Google Cloud Platform that can ping ipv6.google.com successfully. I'm using these GCP docs regarding IPv6 support, including :

• https://cloud.google.com/vpc/docs/vpc#ipv6-addresses
• https://cloud.google.com/vpc/docs/using-vpc#subnet-enable-ipv6
• https://cloud.google.com/compute/docs/ip-addresses/configure-ipv6-address

[NOTE: Edited/updated to provide full commands used]

Here's my simple and repeatable proof of concept:

1. Within a project, create a VPC network

$ gcloud config set project my-test-project

$ gcloud compute networks create targetnet \
   --subnet-mode=custom --mtu=1460 --bgp-routing-mode=regional

2. Create subnet with IPv6 support

$ gcloud compute networks subnets create targetnet-1 \
   --network=targetnet --range=10.9.9.0/24 \
   --stack-type=IPV4_IPV6 --ipv6-access-type=EXTERNAL \
   --region=us-west2

3. Create an instance with IPv6 support

$ gcloud compute instances create test-1 \
   --stack-type=IPV4_IPV6 --ipv6-network-tier=PREMIUM \
   --subnet=targetnet-1 --zone=us-west2-a \
   --image-family=debian-10 --image-project=debian-cloud \
   --machine-type=e2-micro 

4. Add a firewall rule to give SSH access to the instance

$ gcloud compute firewall-rules create target-ssh-home \
   --network targetnet --allow tcp:22 --source-ranges <my IPv4 addr>

5. SSH to the instance, look at the network addresses & route:

$ ip -6 addr
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 state UNKNOWN qlen 1000
    inet6 ::1/128 scope host 
       valid_lft forever preferred_lft forever
2: ens4: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1460 state UP qlen 1000
    inet6 2600:1900:xxxx:xxxx:0:1::/128 scope global 
       valid_lft forever preferred_lft forever
    inet6 fe80::4001:aff:fe04:2d2/64 scope link 
       valid_lft forever preferred_lft forever

$ ip -6 route
::1 dev lo proto kernel metric 256 pref medium
2600:1900:xxxx:xxxx:0:1:: dev ens4 proto kernel metric 256 pref medium
fe80::/64 dev ens4 proto kernel metric 256 pref medium
default via fe80::4001:aff:fe04:201 dev ens4 proto ra metric 1024 expires 85sec pref medium

6. Attempt to ping an external IPv6 resource:

$ ping6 ipv6.google.com
PING ipv6.google.com(lax31s01-in-x0e.1e100.net (2607:f8b0:4007:80e::200e)) 56 data bytes
From fe80::4001:ff:fe00:0%ens4 (fe80::4001:ff:fe00:0%ens4): icmp_seq=1 Destination unreachable: No route

The docs indicate the firewall defaults should allow outgoing ICMP. Still, I've also tried adding firewall rules to allow ICMP (and every) service in various forms to see if that helped. This did not change the result. Also, I confirmed that GCP added a default IPv6 route for the network. (::/0).

According to the documentation, the GCP is to provide a /64 and the NIC should be assigned the first address from that range. I don't see that here. Is there something I need to do on the host to get that? Or is there something I've missed that is needed in GCP?

Solution 1:

Connecting to Google APIs and services from VPCs via external IPv6 addresses is currently not supported. Efforts to ping ‘ipv6.google.com’ from Cloud VM via IPv6 often lead to a destination unreachable ICMP response.
I had replicated your setup with the steps you mentioned and I was also getting the same “Destination unreachable: No route” error while trying to ping ipv6.google.com, however I was able to ping other IPv6 sites like ‘wikipedia.org’ quite successfully.

Furthermore, to best use VMs with IPv6 network, you can prefer incorporating GCP Load Balancing with IPv6 support.