+++ This bug was initially created as a clone of Bug #1792467 +++ Description of problem: DHCPv6 addresses all have a /128 netmask, which breaks the check in baremetal-runtimecfg for whether an IP is in the same subnet as the VIP. This causes deployments using IPv6 to fail when starting mdns-publisher because it doesn't know what address to listen on. In the short term we will just assume /64 for IPv6 addresses, but eventually we need to look up the correct netmask and use that.
Verified on OCP 4.3.0-0.nightly-2020-02-03-115336-ipv6.1 Deployment of cluster with 3 masters and 0 workers succeeded, DHCPv6 addresses has /64 netmask virsh net-dhcp-leases baremetal Expiry Time MAC address Protocol IP address Hostname Client ID or DUID ------------------------------------------------------------------------------------------------------------------- 2020-02-11 09:59:34 52:54:00:54:9c:63 ipv6 fd2e:6f44:5dd8:c956::124/64 provisionhost-0 00:03:00:01:52:54:00:54:9c:63 2020-02-11 10:11:21 52:54:00:85:f1:4a ipv6 fd2e:6f44:5dd8:c956::121/64 master-0 00:03:00:01:52:54:00:85:f1:4a 2020-02-11 10:16:00 52:54:00:93:14:83 ipv6 fd2e:6f44:5dd8:c956::103/64 master-1 00:03:00:01:52:54:00:93:14:83 2020-02-11 10:10:06 52:54:00:e2:f6:a0 ipv6 fd2e:6f44:5dd8:c956::13a/64 master-2 00:03:00:01:52:54:00:e2:f6:a0
Since the problem described in this bug report should be resolved in a recent advisory, it has been closed with a resolution of ERRATA. For information on the advisory, and where to find the updated files, follow the link below. If the solution does not work for you, open a new bug report. https://access.redhat.com/errata/RHBA-2020:0528