1

It's very strange issue, on openstack cluster we have VxLAN between network and compute node and its using 1450 MTU (1500-50) because of encap header. Because of it some of my application not working properly. I am planning to increase MTU (jumbo) 9000 on all my openstack cluster node so question is in that case do i need to increase MTU on all my physical cisco switches? We don't want to do that.

Just wanted to know increase host MTU size required network switch MTU size to?

Reference question i have posted: https://unix.stackexchange.com/questions/322011/multiple-puppetmaster-behind-load-balancer/322126#322126

1
  • Did any answer help you? If so, you should accept the answer so that the question doesn't keep popping up forever, looking for an answer. Alternatively, you can provide your own answer and accept it.
    – Ron Maupin
    Aug 6, 2017 at 22:51

1 Answer 1

3

You haven't described the network very well, but if your switches are not set to use jumbo frames, any jumbo frames reaching a switch will be dropped as a giant.

Jumbo frames are not really standardized. You need to make sure that all the devices through which the jumbo frames pass can handle the maximum MTU of any frames passing through those devices.

6
  • our legacy network running on 1500 which we can't change because of production. Can i change openstack Host MTU to 1550 i believe it will solve my problem but not sure about how lagacy cisco switch will react on 1550 MTU
    – Satish
    Nov 9, 2016 at 18:33
  • Cisco switches default to the standard-defined ethernet 1500 MTU (plus 18 or 22 bytes for the ethernet frame itself). Any frames larger than 1518 (1522 for frames with VLAN tags) will be dropped as giants.
    – Ron Maupin
    Nov 9, 2016 at 18:37
  • Why it will drop instead of doing fragments?
    – Satish
    Nov 9, 2016 at 18:51
  • Ethernet (layer-2) doesn't fragment, that is in IP (layer-3). Ethernet drops and increments the interface error counter.
    – Ron Maupin
    Nov 9, 2016 at 18:53
  • agreed with you.. I can increase puppet size on neutron l3 router so it can fragment packets.. let me think what we can do
    – Satish
    Nov 9, 2016 at 19:02

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.