Hi,

I have RB230 with RB44 router.
I was use it with ROS 2.8.26 until last week.
the network topology as followings:


Eth-1 as Public connected to Cisco Ethernet.
Eth-2 and Eth-3 was bridged and as Local connected to G.SHDSL (CO) Ethernet.
And G.SHDSL (CPE) Ethernet connected to Hub/Switch.
And my computers connected to that hub at my remote site.

All things work fine until i was upgrade to ROS 3.6 with the same config,
We could ping everywhere. All domains resolved fine. Web sites with small
Packets like "google.com" works. But No messenger, no heavy websites such
as msn.com !!!!!
I was changed my MTU of my computers to 1200 with Dr. TCP I was feel better
More sites could be reached but still no messenger and no msn.com.

Then I was found very wonderful thing, when I was ping my computers from the
Router and with packet size grater than 1480 I was got "Pocket to large and cannot be
Fragmented" message and when I was check "don't fragment" tick box and when not
I was got the same message. I was check MTU of my bridge and my Ethernet port all
1500 and then I was change all MTUs to 1480 and noting changed!

All of my settings work fine with ROS 2.8.26

Could you please tell what's wrong?


Best Regards
Ali J. Shirazi

I'm beginning to think my problem is the same. I've gotten rid of OSPF, disabled webproxy, played with srcnat till I'm blue in the face, srcnat seems stable now.

Linux boxes are still randomly freaking out, Windows doesn't have this problem. I haven't been able to get results with the MTU though. Could be every time I check its working properly.

j_shirazi -

I have a couple of ROS 3.6 sitting right here (PC based) and have absolutely no issues when using packet size of 1500.... I can ping anywhere anytime.....

Now - looking at your setup -

You have a Cisco - that is one of your public interfaces.

You also have a G.SHDSL (not sure what that is but I know what a SDSL is - so I am guessing they are very similiar).

Eth-2 and Eth-3 was bridged and as Local connected to G.SHDSL (CO) Ethernet.
And G.SHDSL (CPE) Ethernet connected to Hub/Switch.

Not really sure what you are saying here. I understand that you have eth 2 & 3 bridged together. Then you say 'Local', is that also bridged to eth 2 & 3?

And it looks like you are saying that your computers and clients are connected to the same interface as the G.SHDSL via your hub/switch. Is that correct?

Basically what it looks like to me is that;
1) You and your clients are going out the G.SHDSL interface and not the 'public' interface provided by the Cisco on eth1.

2) G.SHDSL type interfaces typically use PPoE or PPoA protocols, and guess what, those protocols have the MTU set for 1480, the remaing 20 bits are for the overhead the protocol uses....

3) Since you are no longer going out eth1 then your configuration has changed some how - so post it here (minus any public IP addresses that you do not want publically posted here).

4) Also typically - PPoE and PPoA use dynamic IP addresses, and ISPs are notorious for making note of the original interface MAC address connected to them (The PPoE/PPoA interface) and not letting any other MAC address access to that line without a reset at the ISP....

R/

Hi,

Not really sure what you are saying here. I understand that you have eth 2 & 3 bridged together. Then you say 'Local', is that also bridged to eth 2 & 3?

Eth 2 & 3 bridged as local ( together ).one port connected to G.SHDSL and one local computer with same rules.

1) You and your clients are going out the G.SHDSL interface and not the 'public' interface provided by the Cisco on eth1.

2) G.SHDSL type interfaces typically use PPoE or PPoA protocols, and guess what, those protocols have the MTU set for 1480, the remaing 20 bits are for the overhead the protocol uses....

3) Since you are no longer going out eth1 then your configuration has changed some how - so post it here (minus any public IP addresses that you do not want publically posted here).

4) Also typically - PPoE and PPoA use dynamic IP addresses, and ISPs are notorious for making note of the original interface MAC address connected to them (The PPoE/PPoA interface) and not letting any other MAC address access to that line without a reset at the ISP....

My clients data pass through the modem (work as bridge).No PPPoE and PPPoA.

This topology works good on 2.8.26 ROS. and fails on 3.6 ROS.


Regards
Ali

Hi,

I guess i was found the problem finally !


The problem is the bridge. All MTU of the bridge members different from the others
This problem was happened. thats Funny this problem not existed on 2.8.x version!
Thats really bad problem .


Ali