cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
3731
Views
30
Helpful
2
Replies

SDA IP Directed Broadcast feature

JP10
Level 1
Level 1

Hello community,

 

Recently we are struggling with silent hosts in our Fabric domain. Many of our end devices in VN are either remote management cards of different vendors or management interfaces of proprietary server solutions. Both are acting as silent hosts, so we have an issue onboard them and keep them alive. The count has already reached over 150 hosts and simultaneously we're building additional VNs for a different purpose, which has an impact on hosts making them unreachable again until ping VRF from edge switch. 

 

A few days ago we discovered the IP Directed Broadcast feature available from version 2.1.2 (running 1.3.3.7) and considering this upgrade together with IOS to a supported version of this feature. Does anyone have more insight into how this feature works? Could this be the fix for a workaround with static device-tracking binding on specific Fabric Edge / interface while trying to reach silent hosts from outside the Fabric? 

 

Thank you in advance, every help will be appreciated. 

1 Accepted Solution

Accepted Solutions

Scott Hodgdon
Cisco Employee
Cisco Employee

JP10,

Yes, the new IP Directed Broadcast feature should help with this. Here is a description of what happens in your use case:

Feature Description

A Catalyst 9000 Series SD-Access Border switch can convert an IP-directed broadcast into an Ethernet broadcast and flood to all endpoints in the destination VLAN.

 

Silent Hosts

An endpoint whose location in fabric is not known because it has not sent any packets or frames.

The Border Node will ARP for the host if the destination IP address is not in LISP map-cache or in ARP cache.

Cheers,
Scott Hodgdon

Senior Technical Marketing Engineer

Enterprise Networking and Cloud Group

View solution in original post

2 Replies 2

Scott Hodgdon
Cisco Employee
Cisco Employee

JP10,

Yes, the new IP Directed Broadcast feature should help with this. Here is a description of what happens in your use case:

Feature Description

A Catalyst 9000 Series SD-Access Border switch can convert an IP-directed broadcast into an Ethernet broadcast and flood to all endpoints in the destination VLAN.

 

Silent Hosts

An endpoint whose location in fabric is not known because it has not sent any packets or frames.

The Border Node will ARP for the host if the destination IP address is not in LISP map-cache or in ARP cache.

Cheers,
Scott Hodgdon

Senior Technical Marketing Engineer

Enterprise Networking and Cloud Group

anthony.wild
Level 1
Level 1

Not to drag up an old thread, but if its a very very minor subset of devices, have you tried enabling NTP on the endpoint? We had a small minority of devices (like 10 of 5000) endpoints that had this issue... mainly door access reader panels. We enabled NTP on them just to make them inherently chatty and that solved the issue. I'm not sure what the smaller lift here is in terms of overhead on the fabric but wanted to make mention.