cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
2363
Views
0
Helpful
5
Replies

MTU fragmentation 3750 switch

tedauction
Level 1
Level 1

Hello, we have two servers on the same subnet which are having trouble transferring data via UDP between themselves using proprietary software. For instance, backups will partially complete, then fail and have to repeat the backup.

I am thinking that possibly MTU fragmentation is occurring. The Cisco 3750 interfaces are set for normal 1500 bytes MTU. Perhaps NICs on the servers is set to a higher value ?

Should I simply increase the MTU value on the switch interfaces connecting the two servers to a higher value e.g. 2000 or 2500 ?

 

Thank you for any advice.

1 Accepted Solution

Accepted Solutions

Reza Sharifi
Hall of Fame
Hall of Fame

Hi,

Since the servers are in the same layer-2 vlan, there will be no fragmentation. On the 3750 series switches, the mtu change is global, meaning you have to enable it for all ports.  Also, after enabling it you have to reboot the switch in order for the change to take effect. If you do it on the switch side, you would need to make sure the servers are also configured with larger mtu. You can simply raise it to 9216 on both the switch and the server sides and test again.

HTH

View solution in original post

5 Replies 5

chrihussey
VIP Alumni
VIP Alumni

Hello,

One would assume an application and standard protocols should have some form MTU discovery built in. However, depending on the interfaces involved (Fast Ethernet vs Gig), you can adjust to different settings. You may also have to reload the switch for it to take effect. Reference the link below, it may not be your particular model, but it's probably consistent across the family:

 

https://www.cisco.com/c/en/us/td/docs/switches/lan/catalyst3750x_3560x/software/release/15-2_4_e/configurationguide/b_1524e_consolidated_3750x_3560x_cg/b_1524e_consolidated_3750x_3560x_cg_chapter_010000.pdf

 

One final thought. Is there a chance you could connect the servers back to back, eliminate the switches and run the application? If you got the same results, you could rule out the switches and focus on the real issue.

Hope this is of some help.

Reza Sharifi
Hall of Fame
Hall of Fame

Hi,

Since the servers are in the same layer-2 vlan, there will be no fragmentation. On the 3750 series switches, the mtu change is global, meaning you have to enable it for all ports.  Also, after enabling it you have to reboot the switch in order for the change to take effect. If you do it on the switch side, you would need to make sure the servers are also configured with larger mtu. You can simply raise it to 9216 on both the switch and the server sides and test again.

HTH

Even though the servers are on the same VLAN, they are ESX host VMs, therefore I was thinking possibly the virtual NICs on the VMs might have MTU settings higher than my Cisco switch e.g. if the VM NIC had an MTU of 2000 and my Cisco switch has 1500, then there would be fragmentation when the packets reached my Cisco switch ?

Thank you for clarification.

Since they are ESXi hosts, they may have a different MTU setting specially if there is also storage involved.

Try to match on both ESXi hosts as well the switch and test again. Make sure to check the storage heads as well.

HTH

nader.zaman
Level 1
Level 1

1) Are the servers connected to the same switch?

2) Are you seeing any errors on the switch interfaces?

3) Can you post result of "show interface ...." for each server?

4) Are the switch interfaces and server NICs both configured to auto negotiate for speed and duplex?

 

What you describing is not a result of L2 fragmentation. There is something else wrong.

 

 

 

 

Getting Started

Find answers to your questions by entering keywords or phrases in the Search bar above. New here? Use these resources to familiarize yourself with the community:

Innovations in Cisco Full Stack Observability - A new webinar from Cisco