Open allenrobel opened 3 years ago
One followup on this. For the use-case I was working with (a fabric based on RFC5549 and using BFD) the DUTs are configured with only IPv6 BGP sessions (IPv4 address-family runs over the IPv6 session and interfaces are configured with ip forward
+ IPv6 address). I was initially thinking that no ipv6 redirects
was required for the BGP sessions to stabilize due to BFD being used for the IPv6 BGP sessions. And, no ipv6 redirects
DOES work to stabilize these BGP sessions.
But I just tested removing no ipv6 redirects
and adding only no ip redirects
and this also stabilizes the IPv6 BGP sessions.
So, for this specific use-case, NX-OS Ansible support for ipv6-specific handling of redirects is not actually needed. Though, it would probably still be good to have.
SUMMARY
The current YAML structure for nxos_l3_interfaces doesn't seem to provide the ability to configure ipv6 redirects or unreachables. Since both
redirects
andunreachables
parameters reside directly undername
and seem only to configure ipv4 (i.e.[no] ip redirects
and[no] ip unreachables
)ISSUE TYPE
COMPONENT NAME
nxos_l3_interfaces
ADDITIONAL INFORMATION
The obvious fix would be to move
redirects
andunreachables
into the ipv4 dictionary along withaddress
. But this would break existing playbooks.Perhaps existing ipv4
redirects
andunreachables
(when appearing directly undername
) could be deprecated, while simultaneous support is added for these same parameters within the respective ipv4 and ipv6 dictionaries? Legacy usage could be maintained for some time, parallel with new support being added for these parameters under the ipv4/ipv6 dictionaries? And any future protocol-specific parameters could then be added within these existing dictionaries.A less desireable option would be to add separate
ipv6_redirects
andipv6_unreachables
parameters (though, future users will likely scratch their heads wondering why these weren't added to the ipv6 dictionary from the start...)Below is current usage:
Below is the proposed change (showing support for legacy (deprecated) ipv4 usage on Ethernet1/4 while simultaneously adding support for ipv6 redirects and unreachables within the ipv6 dictionary. Ethernet1/6 shows proposed new usage for both ipv4 and ipv6).