Reading a doc from Juniper, Clos IP Fabrics with QFX5100 Switches. I saw ebgp export policy with next-hop self. And the explanation is "we use next-hop self to change the next hop of each advertised prefix to the egress interface of the switch." As I know this is ebgp default behavior.
Appreciate if someone give more detail explanation on this.
One more question about bgp graceful-restart, in the same doc, "graceful-restart—Of course we want the ability to make policy changes to BGP without having to tear down existing sessions. To enable this functionality, we can enable the graceful-restart feature in Junos OS" is it true? is it bgp graceful-restart functionality?
Probably you can point to the document, it will be helpful to understand why it has been mentioned that way or if document needs correction. Your understanding is correct by default, on a e-bgp peering next-hop self is changed automatically.
Graceful-restart is the capability to take out router peering and re-establishing minizing distruption to other part of the network. Helper mode for graceful-restart (to assist peer node going down) is enabled by default.
Ideally changing BGP policy, should never cause peering to go down. However, if the policy is moved from peer level to group level or vice-versa it may cause peering to go down. This is different than graceful-restart functionality and the two should not be confused.
Thanks for your reply. I still want to confirm if my knowledge is out of date or the doc need correction.
The doc link is https://www.juniper.net/us/en/local/pdf/whitepapers/2000565-en.pdf, page 14&15
As Kingsman mentioned, there is no requirement to exclusively to do NHS for the ebgp sessions.