Does Jumbo Frames ever get fragmented

[shiraazs]

Hi

some questions regarding jumbo frames.

1) If the switch is NOT configured for jumbo frame support does the frame always get dropped or are there conditions under which the frame gets fragmented.

2) If they are always dropped then which type of frames get fragmented ?

3) If jumbo frames support is configured on switch but the egress port towards an 1841 router is 100MB , I understand the frames will get dropped.  Is there any way to solve this issue. If the link to the router is upgraded to gigabit will the frame pass thru the egress port to the router ? will the router fragement and route it onward ?

Thanks

[Joseph W. Doherty]

Disclaimer

The   Author of this posting offers the information contained within this   posting without consideration and with the reader's understanding that   there's no implied or expressed suitability or fitness for any purpose.   Information provided is for informational purposes only and should not   be construed as rendering professional advice of any kind. Usage of  this  posting's information is solely at reader's own risk.

Liability Disclaimer

In   no event shall Author be liable for any damages whatsoever (including,   without limitation, damages for loss of use, data or profit) arising  out  of the use or inability to use the posting's information even if  Author  has been advised of the possibility of such damage.

Posting

My understanding, L2 receiver has to be able to handle the frame size the sender xmits or it will drop it.

If the device supports different frame sizes on different interfaces, larger L3 can be fragmented into smaller at L3; don't know of any L2 fragmentation.

So, for example for your last question, if the router can accept the jumbo ingress frame, if there's L3 boundary (normal on a router), L3 can fragment the packet into the egress interface's smaller (new) frames (not fragmented original frames).  However, if this was a L3 switch that supports different interface frame sizes, and it was just a L2 hop, the egress interface, I believe, would drop the ingress interface's frame.

[shiraazs]

Hi

Thank you for your reply. So if I understand correctly if a packet is switched (L2) it will be dropped and not fragmented if the packet is bigger that the configured MTU however if teh packet is routed (L3) it will get fragmented ?

[william.briere]

The layer 3 packet would be fragmented unless don't fragment (DF) is set.

[rsimoni · Accepted solution]

Hi Shiraaz,

fragmentation occurs at L3 not a L2 (unless we deal with FR or ATM or other technologies). For sure it does not apply to ethernet.

Other important concept is that the MTU MUST be configured with the same value on each L3 link between 2 routers.

router1 1500 --------- 1500 router2 9000 ------- 9000 router3 9000 ---------- 9000 router4 1500 --------- 1500 router5

fragmentation will occur between router4 and router5 assuming that router4 gets a frame with size 9000 from router3 and will fragment in at least 6 smaller 1500 frames to be sent to router5 (this is a scholastic excercise as since our ip packet most likely is coming from router1 its size won't be larger than 1500).

if you have a MTU mismatch between 2 ends of a link you will have drops in the larger-to-smaller direction

router1 9000 ------- 1500 router2

router2 will drop all the frames bigger than 1500 (well a little bit more) coming from router1

this scenario is clearly a configuration mistake!!!

if you enable jumbo support on one link you need to be sure that on the other side of it there is a l3 device which also supports jumbo frames

L2 switches simply drop frames bigger than the port mtu.. but if this happen you clearly configured something wrong.

Say that you set a L2 port of a L2 switch to 9000, since the mtu is calculated at l3 the real frame size is the one calculated by the L3 device (host or router) which sends the frame, therefore most likely will be 1500. So there is no risk to drop anything on the switch.

hope this clarifies

Riccardo

[Joseph W. Doherty]

Disclaimer

The    Author of this posting offers the information contained within this    posting without consideration and with the reader's understanding that    there's no implied or expressed suitability or fitness for any  purpose.   Information provided is for informational purposes only and  should not   be construed as rendering professional advice of any kind.  Usage of  this  posting's information is solely at reader's own risk.

Liability Disclaimer

In    no event shall Author be liable for any damages whatsoever  (including,   without limitation, damages for loss of use, data or  profit) arising  out  of the use or inability to use the posting's  information even if  Author  has been advised of the possibility of such  damage.

Posting

SHIRAAZ SINGH wrote:

Hi

Thank you for your reply. So if I understand correctly if a packet is switched (L2) it will be dropped and not fragmented if the packet is bigger that the configured MTU however if teh packet is routed (L3) it will get fragmented ?

Yes, although as William noted (with IP as L3) DF bit set should stop fragmentation (if DF is honored).

[shiraazs]

that clears up the mystery. Thanks guys for your help and for taking the time to reply. Makes a whole lot of sense now !

[rsimoni]

Hi Shiraaz,

glad we could help.

would you mind rating the helpful answers and close the thread if you don't have any other questions on the topic?

R

[shiraazs]

Just one more question please. I have the following:

SWITCH A ---------------------- Cisco 1921 router A --------- < wan link > ------------------- Cisco 1921 Router B -------------- SWITCH B

- The Switches have Jumbo frames enabled on the ports towards the routers

- The switches are HP Procurve so Jumbo is enabled at VLAN level and not at Port level

- The routers can only handle MTU of 1500

- The routers are routing between Lan and Wan

- The ports are at gigabit , full duplex.

- On the ends are HP Lefthand Storage Nodes configured for Jumbo Frames

- The storage nodes are in the same VLANs as the switch and router LAN ports

- The Storage nodes need to replicate to each other

1) will the jumbo frame pass from the switches to the routers or will they be dropped ?

2) If they are passed , will the router fragment the packet and route it onward ?

3) if point 2 is true , will the packet make it from storage node A side to storage node B side successfully and vice versa ?

4) If point 1 is true , any suggestions on how to solve the problem . Removing Jumbo frames completely is not an option at  

    this time . Customer dictation

Thanks for the answers and advice

[rsimoni · Accepted solution]

I see issues with your setups.

you have a MTU mismatch between the Hp storage and the 1921 (since they are in the same vlan I suppose that the 1921 is the default gateway for the storage devices). The MTU check will fail for packets bigger than 1500 received from the HP devices on the 1921 with consequent drop.

Customer dications like the one you wrote are simply non sense and really make me laugh. They are dictated by people who clearly don't understand networking world.... it is like tomorrow morning I wake up and decide that you can walk only using your right foot!!! clearly a non sense!

Solution might be to make the switch A - 1921 a L3 link (routed link) with MTU 1500 . This way the Hp switch will fragment frames in the 9000 to 1500 path but at least there will be no drop.... Of course the easiest way would be to reduce the MTU on HP devices to 1500 and keep everything connected at L2 (3 devices in the same vlan).

[shiraazs]

Thanks Ricardo

The  D/gateway for the storage devices is actually the HP switch and not the 1921 but I think the problem remains the same.

Thanks again for your help. Will close the thread now

[rsimoni]

ok I see.. so the hp sw is the dg but its next hop is a device in the same subnet (the 1921). In terms of MTU it does not change a thing as what I wrote above still applies (downside is that the hp switch will generate lots of icmp redirect to the server, unless they are disabled.. but this is not part of this problem)

good luck with your attempt to explain this to the customer 

[Kishore Chennupati]

oh customers and their requirements

[casey454]

rsimoni are you really a Cisco employee? Seems like you could have just explained that all switchports in the layer2 path have to have Jumbo enabled for it to work. 

You could have also just explained how fragmentation works at layer3 interfaces and the fact that jumbo frames would affect that. What mitigation techniques would remedy this like tcp mss-adjust.

You're not impressing anyone with your knowledge here. Also, you end with advice to disable Jumbo frames?