Red Hat Bugzilla – Bug 1475983
[RFE] Please backport support for AdvRASrcAddress (selection of RA source address)
Last modified: 2018-10-30 03:49:09 EDT
Description of problem: Currently the shipped radvd version does not support AdvRASrcAddress (that's selection of RA source address). This is required for fully redundant setups, e.g. with two routers, keepalived and VRRP while having only one client that has 2 uplinks. A more complete description of the specific scenario is shown at https://github.com/reubenhwk/radvd/issues/45#issuecomment-103113580 as a visualized example. https://github.com/robbat2/radvd/commit/9aad6f02decbb65946121aef9347c17f659124b0 is the fully working upstream implementation that was done for radvd 2.16, so please backport it to RHEL (either the feature itself or rebase to >= 2.16). Configuration example for /etc/radvd.conf (same file on both routers) --- snipp --- interface eth1 { AdvSendAdvert on; AdvRASrcAddress { fe80::1; }; prefix 2001:db8:1::1/64 { }; RDNSS 2001:db8:1::1 { }; }; --- snapp --- Configuration example for /etc/keepalived/keepalived.conf (same relevant part on both routers): --- snipp --- vrrp_instance VRRP_INSTANCE { # [...] virtual_ipaddress { 2001:db8:0::1/64 dev eth1 fe80::1/64 dev eth1 } # [...] } --- snapp --- What happens without this AdvRASrcAddress? The client gets two default routes, one like fe80::router1 and one like fe80::router2. However with keepalived and VRRP, only one (either router1 or router2) has a globally routeable IPv6 address, thus the client might be "offline" depending on the chosen default route. Playing with settings like AdvDefaultLifetime or AdvDefaultPreference does not make sense, because they are too slow to be useful for quick keepalived takeovers (and fallbacks). Version-Release number of selected component (if applicable): radvd-1.9.2-9.el7 How reproducible: See above and below. Actual results: No support for AdvRASrcAddress (selection of RA source address) Expected results: Support for AdvRASrcAddress (selection of RA source address)
Cross-filed ticket 01899369 on the Red Hat customer portal.
Bah, the global unique IPv6 addresses should be indeed all exactly the same one, my fault when obfuscating from real-world configuration...
Our tests using radvd-2.17-1.el7.cs01899369.x86_64 (via GSS) looked good :)
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-2018:3027