Upon getting decided MTU values for a given path (see Maverick’s reply), it is perhaps price searching for traces of TCP MSS clamping (a.okay.a ip tcp adjust-mss in Cisco communicate) taking place alongside the trail between supply an vacation spot.
Why? If the assessments occur to be TCP based mostly, outcomes is perhaps skewed by MSS clamping.
In a typical situation on plain Ethernet, an initiator (largely referred to as ‘consumer’, however TCP has no idea of consumer nor server), in its intial TCP SYN phase, will announce a TCP Most Phase Dimension (MSS) of own-NIC’s-MTU_minus_(TCP+IPv4-Headers), or 1500 – 40 = 1460 (1428 if TCP timestamps are in use).
And the responder (‘server’) will do the identical, deriving the utmost TCP payload dimension from it is personal NIC and sending that worth in it is SYN/ACK phase.
Routers, alongside the trail, interconnecting totally different hyperlinks/media with totally different MTU traits, can and can manipulate that worth, in order that fragmentation may be prevented. This turns into crucial if additional encapsulation is occurring on some sections of the trail, like PPPoE, GRE or IPSec tunnels.
To watch the results of TCP MSS clamping, you want to have the ability to seize the preliminary SYN and SYN/ACK packets on each initiatior and responder for a TCP session.
If, in the identical SYN phase, upon arrival on the responder, the worth is decrease than it was when it left the receiver, then at the very least one router (or L3-switch or firewall or VPN concentrator or…) is doing TCP MSS clamping.
Make sure to examine the values in each instructions (with the SYN/ACK packet of the identical tcp session). PathMTU in itself is one thing unidirectional that happens independently within the different route, and in addition the MSS clamping configuration bits on a given system may be of unidirectional nature (solely ingress, solely egress, each) – relying on vendor, that’s.
UDP has no idea of MSS; both the appliance is aware of what number of bytes to ship per packet, or it depends on PMTU discovery and it is hints of decrease MTU, and adapts accordingly. However PMTU on IPv4 just isn’t particularly well-known for being dependable.
Out of expertise, if I had been requested (by software folks, that’s) about “do we’ve got 1400bytes MTU on our community?”, I’d bounce the query again: “1400 is what you want? If I affirm, will you be proud of sending a most of 1360bytes of payload per packet on non-timestamped TCP and 1372 on UDP?” In the long run, I’d need them to carry forth their necessities when it comes to n bytes of payload per packet, not as MTU or packet dimension or no matter they imagine the unit is.