IPv6 VIP addresses need to be /128
authorMichele Baldessari <michele@acksyn.org>
Wed, 26 Apr 2017 09:40:51 +0000 (11:40 +0200)
committerMichele Baldessari <michele@acksyn.org>
Wed, 3 May 2017 10:32:09 +0000 (12:32 +0200)
commit6227484b60cd72cf4647051923a3baf175100a72
tree12670c866dda35c3517cd2bcbfc3b2a727a4090b
parentbaecff6c3ca47400ace7fb6ef5a83cd0a206041a
IPv6 VIP addresses need to be /128

We currently hardcode /64 as our VIP addresses when using IPv6.
The problem with this is that some server code might bind to that
IP as a source address when doing inter-cluster communication
(rabbitmq/galera for example). So when the VIP moves there will
be effectively a network outage between the nodes, which should not
happen.

Likely this was hardcoded to /64 because the RA IPaddr2 needs a nic
parameter when /128 is specified. This is due to:
https://bugzilla.redhat.com/show_bug.cgi?id=1445628

We also make sure we use the ipv6_addrlabel option set to 99 so that
they will never be used as source ip addresses.

Depends-On: I7fcf15a00aedbdcfb21db501ad46c69fb97ec30c
Partial-Bug: #1686357

Change-Id: Ibefde870512ad1e03ff12f7aea91b3734f03f96f
Co-Authored-By: Sofer Athlan-Guyot <sathlang@redhat.com>
Co-Authored-By: Marios Andreou <mandreou@redhat.com>
Co-Authored-By: Damien Ciabrini <dciabrin@redhat.com>
manifests/pacemaker/haproxy_with_vip.pp