VMware NSX

I have a lab with vtep implemented, but with the ping tests, the icmp packets reach a maximum of 1472 bytes. It occurs that the networks are extended and operating in my labor.

My question is, the MTU 1600 a requirement for extended network operation or is it just a performance issue or something in this sense?

Below is the recommendation from VMware regarding the MTU sizing

Your normal ping packet size is between 32 ~ 56 bytes depending on the type of O.S, so you don't observe any issue.

During normal data transfer, the packet size will be 1500 bytes. To this additional bytes will be added to the header. (VXLAN header)

So its recommended to have the MTU size to 1600 bytes to avoid any packets getting dropped.

Thanks rajeevsrikant, but i still have the same question.

Did you inform us that VMWARE "recommends" ... that is, is it a recommendation or requirement?

I'm testing with the command below:

vmkping ++netstack=vxlan -s 1472 -d -I vmk1 ip_dst (works fine!)

vmkping ++netstack=vxlan -s 1600 -d -I vmk1 ip_dst (dont work!)

When the packet size exceeds 1472 bytes, ping does not work, however my extended network is working ... so I wonder if it's a performance issue ...

thx!

You must use every time -28 bytes ( -8 bytes ICMP header and -20 bytes for IP ) so correct is 1572. ( 1600 MTU )

i get it...  thnx

but on a network where I do not have the MTU of 1600 bytes, will vxlan work? Or the vxlan will run with a problems?

thanks so much!

VXLAN overhead 50 bytes. You need to set recommend minimum by VMware 1600 MTU or higher.  Total = 1554 bytes

novakk_AH  if don't i do this? Does vtep work? can the vxlan package be fragmented?

Sorry for my insistence, I'm just trying to understand this better ...

thanks so much!

The main point to consider here is that the VXLAN packets cannot be fragmented.

That is the reason always ensure that the MTU is set to 1600 bytes.

\O tnx so much my friend!

Rajeev is correct, 1600 is required, and to add since you asked, it will be end to end from VTEP to VTEP and everything in between.