EXOS: missing RSPAN data because of HW limitations (GRE MTU bigger than connection MTU)

[Volker_Kull]

Using RSPAN feature on EXOS HW there are situations where the recieved data on the remote RSPAN device is not complete. This occurs when the RSPAN process (RSPAN sender) on EXOS HW should transport a 1500 byte packet via GRE-Tunneling to the RSPAN receiver and the encapsulated packet (1500 byte + GRE header) is too big for the available MTU between RSPAN sender and receiver.

In this case you will see at the RSPAN receiver only packets matching the connection MTU. There is no note/description or error mesage available about this silent drop because of HW limitations. We figured that out by comparing local mirroring data with remote mirroring (RSPAN) data.

Another solution is to configure Jumbo frames in the complete communication path betwee RSPAN senden and receiver. This is sometimes possible in local connections but not in big LANs and via WAN. 

For that we need

• a log message and/or a counter for drops in active RSPAN connections.
• possibility of truncation of the data that the RSPAN packet matches the connection MTU to RSPAN receiver

Br

Volker

[AlexN]

We now have another customer who needs it, thus it qualifies for FR.
Feature Request 02334617 Ability to set an IP MTU and MSS value on GRE tunnels

[AlexN]

Can’t agree more. 

What if we use Fabric as a transport ? Will it save your world ? 🙂

(truncated frames need CPU processing if you do it on switch,  I am afraid)

[Volker_Kull]

Did you install IAH? I`m running FGT and other stuff on it. This is not usable in the field and much too expensive for a part time usage. And what´s about VSPs without IAH?

RSPAN to a remote PC with wireshark is absolutely needed. With truncation and/or jumbo frames - does not matter at all. Best with fragmentation in next HW versions.

Managing worldwide networks you will need RSPAN anytime anywhere…

Volker

[AlexN]

Sounds like a kludge for RSPAN.. And what if we rather use IAH VM to do fragmentation/defragmentation ?

Or simply use remote pcap on IAH VM ?