hello
just for reference i'm asking MT support to answer me about this things:
in RB winbox interface:
L2MTU is full L2 ethernet frame with data accepted from upper layers ? It is unchangeable option in interface menu. Is it active by default and is accepting this maximum frames on interface?

MTU is maximum packet size accepted from upper layers to be inserted in L2 Ethernet frame ? by default it's 1500 everywhere. MPLS/VPLS is not included in this option but in L2MTU size, is it?

BT packet size - is it generating full frame L2 frame (ip data)+(ethernet frame) or it is only L3 ip packet ?

how can you describe me next thing:

(493ah) <-> (450g) <-> (3550) <-> (450g) <-> (493ah)
I ran BT from 493ah to other one on other side (both connected with eth1, which has max 1526 L2MTU).
1. changed MTU to 1524 on those 493ah's eth1 interfaces, for traffic not to be fragmented.
2. ran UDP BT from 493ah to other one, with 1524 packet size. but didn't changed anything on 450Gs
Result: no fragmentation occured on 450gs. max L2mtu is 1524 on them. MTU remained 1500.
So how 450g transported this frames which have size 1524 and is maximum L2MTU for 450g ? where ethernet header is placed ? after 1524 ? I dont think so. So, from this example i can say that BT generates packets with already included standard ethernet header ?

3. finally when i rised MTU on 493ahs up to 1526. and ran test with packet size of 1526.......nothing worked ....450g didn't receive traffic...because it's L2mtu max can receive 1524.......again how did 493 transported packet with size of 1526 when it's L2MTU is max of 1526..so out of here i can say that BT generates full L2 data frames with included standard ethernet headers...am i right ?


sorry for such annoying request. but it's very important ...

BTW: change list of max L2MTUs for RBs on wiki page................. 450G doesn't has 1526 on eth1 !!!!!!!!!! all interfaces have max l2mtu of 1524 !!!!!!!!!!!!! this thing very upsets !!!!!!! why 450g shouldn't have at least one interface with 1526.....
other thing which upsets and frustrates me too much is that Mikrotik can't deal with MPLS+VPLS without fragmentation ((((

Ethernet header is not part of L2MTU, this picture explains everything
http://wiki.mikrotik.com/wiki/Manual:Ma ... n_RouterOS

can i get some discussion regarding my questions right here?
i saw this picture and read this wiki hundred times.

so you say that ethernet headers are not included in this interfaces options and are added to frame by default......
and l2mtu means amount of data with VLAN, MPLS, VPLS only ?

MTU means only IP data ? or i have to rise this changeable option in cases when i don't want to fragment MPLS packets with data ?
or as i understand L2MTU is default amount of data "L2" data enabled on interface and MPLS/VPLS isn't fragmented in case when value on interface is enough ? am i right ?


so when i say VPLS L2MTU is 1526 i have to grab away 14 bytes of ethernet header out of there....and it becomes 1512 ??????????? so my VPLS traffic shouldn't be fragmented on any routerboard with smalletst L2MTU of 1522 ???????????????????? (please answer on this. it's critically important) ....... please.....or i've to rise MTU to maximum of L2mtu to forward VPLS L2 packets without fragmentation ????

In case of VPLS, tunnel is encapsulating full frame which means, to forward vpls packet without fragmentation L2MTU should be:

MPLS:4+MPLS:4+VPLS:4+ETH:14+IP:20+Data:1480 = 1526

1. So, in routerboard interfaces information L2MTU option means "FULL FRAME L2 MTU" ???? maximum L2 packet size with all headers included Ethernet and etc ?

2. And changeable option MTU in interface options means IP data ????? (1480+20) without L2 headers like VLAN MPLS VPLS ETH. ?????????

3. So again BT generates full frame L2 packets with ETH header already inluded so it can pass data without fragmentation ?

Please someone, answer me on this specific questions. Like 1... 2... 3...

In case of VPLS, tunnel is encapsulating full frame which means, to forward vpls packet without fragmentation L2MTU should be:

MPLS:4+MPLS:4+VPLS:4+ETH:14+IP:20+Data:1480 = 1526

1. As from the picture in wiki :
* Full frame means complete packet with ethernet header
* L2MTU - packet without ethernet header IP + vlans + mpls
* MTU - Layer3 MTU this value includes only IP header and payload

2. Also from the first point configurable MTU is layer3 MTU (only IP header + payload)

3. Configurable parameters remote-udp-tx-size and local-udp-tx-size specifies Layer3 MTU (IP header+ payload).

so for reference.... full frame on any routerboard on its interface is L2MTU + Ethernet 14 (no fcs, as wiki says it's being discarded)
for example on 450g L2MTU 1524 + ETH 14 = 1538 bytes
So BT generates packet size of 1524, takes 24 bytes from VLAN MPLS headers or whatever else. then interface adds 14 bytes MAC layer info. and passes it to 450g.......450g passes it further because of destination mac address isn't its own. But if this BT would be designated to this 450g and i haven't changed MTU on it up to 1524, this packet would be discarded. Am i right ?

And fragmentation happens only on L3.... so when L2 450g receives packet higher than his L2mtu it just discards it and can't fragment, yeap ?

1. As from the picture in wiki :
* Full frame means complete packet with ethernet header
* L2MTU - packet without ethernet header IP + vlans + mpls
* MTU - Layer3 MTU this value includes only IP header and payload

2. Also from the first point configurable MTU is layer3 MTU (only IP header + payload)

3. Configurable parameters remote-udp-tx-size and local-udp-tx-size specifies Layer3 MTU (IP header+ payload).

Yes, If you leave MTU to 1500 then BW test will be able to send out only 1500 byte packets.

Ip fragmentation happens only in L3, but, for example MPLS and EoIP can also fragment packets and end customer won't even notice that packet is being fragmented.

It is good behaviour to increase mtu and l2mtu on interface to max values?
Or it can get to an issue?

You should NOT increase the MTU, unless you really need to route jumbo packets. If you have a PMTUD problem somewhere along the path, it will cause MTU issues.

L2MTU you can increase as much as you need without any issues. Unless you need it tho, it will not make a difference.
For example, if the interface is running plain TCP/IP and nothing else, increasing L2MTU to 9k will do nothing.

You should NOT increase the MTU, unless you really need to route jumbo packets. If you have a PMTUD problem somewhere along the path, it will cause MTU issues.

L2MTU you can increase as much as you need without any issues. Unless you need it tho, it will not make a difference.
For example, if the interface is running plain TCP/IP and nothing else, increasing L2MTU to 9k will do nothing.

Ok in case I have some eoip tunnel or MPLS/VPLS interface and some vlan I have to modify only the L2mtu without change MTU, because are all protocol that use L2 stack..exactly?
In what case is necessary to increase MTU??Can you show me an example?
Many thanks

Ok in case I have some eoip tunnel or MPLS/VPLS interface and some vlan I have to modify only the L2mtu without change MTU, because are all protocol that use L2 stack..exactly?
In what case is necessary to increase MTU??Can you ask me an example?
Many thanks

Here is a presentation that deals with MTU when using MPLS/VPLS: http://tiktube.com/video/KHhE3aEKdDoDEJ ... sDlIoFqoq=
After watching that, if you still have any questions, ask.

As for increasing MTU, it is only usefull when you need to route packets larger then 1500 on L3. Meaning, if you need to route jumbo packets.
Usually, jumbos do not leave the L2 segment, meaning communication between 2 hosts with jumbos is usually direct host-to-host traffic, without being routed.

There is nothing preventing you from routing jumbos if you need it tho. As mentioned, watch out for MTU issues over longer routed paths, and make sure PMTUD works, or use other work-arounds, such as changing MSS.
I have not yet seen a network that routes jumbos, but there are valid reasons and scenarios to do so.

Many thanks, I'll watch your presentation!
There is a karma for you!!

Many thanks, I'll watch your presentation!
There is a karma for you!!

Thanks!

If you only care about the MTU part, that starts at 10minutes into the presentation, and then there is some more at the end around 37minutes in.

Hello
I have a CCR 1009 and one 1100AHx2 directly connected with a Cat 6 SSTP cable 1gbit link.

CCR port 8, 1100AHx2 port 13.

Do I achieve any benefits setting MTU and L2 MTU at 9116 (maximum equal size supported by both routers).

thank you