I just recently learned that the BGP Soft Reset feature in Cisco IOS is automatically implemented in software release 12.0(2)S and later. Earlier software releases required the neighbor soft-reconfiguration in the BGP configuration to dynamically update BGP route-maps, local preference, etc. Without the neighbor soft-configuration enabled any configuration changes required a hard reset of the BGP peer which would interrupt network traffic. There was a memory penalty paid to having the neighbor soft-reconfiguration enabled since the router would keep a duplicate copy of the BGP route table in memory;
Before the BGP Soft Reset Enhancement feature, a soft reset for inbound routing table updates was performed by entering the neighbor soft-reconfiguration router configuration command. This command was used to configure the local BGP router to store all received (inbound) routing policy updates. However, this method uses too much memory because inbound updates are not modified and is not recommended.
I’m guessing this new feature had a significant impact for anyone taking a full Internet BGP table?
Cheers!
Image Credit: Earth 3D by Jan
This is true and it’s very handy. However, it’s not as good (in some ways) as soft reconfiguration. Having that local copy of the received table means that I can see everything sent to me even if it was filtered. With the automatic route refresh function, you don’t have that capability any more. For full route tables, that’s definitely of benefit. For smaller tables, I still use soft reconfiguration because the troubleshooting value of that additional copy of the table is enormous in my opinion
Thanks for the clarification John. I was wondering if there were any downsides to the feature. Would I be correct in assuming that you can still use the legacy method by using the neighbor soft-configuration command?
Thanks for the comment!
Correct; soft-reconfiguration inbound still works just as before. From an SP perspective, I tend to see it enabled still on managed CEs for visibility purposes; less so on the PEs where it may be somewhat less scalable. Don’t get me wrong, having route refresh capability available on most peers by default is wonderful though!
Thanks for the education John!