Hello all
Took Router 2011 and put into production as a bridge (tower 7 Rocket M5 and a RB912 link), but my colleagues have shared that there are problems like strange lag and ping tests with big packets (65000) don’t run on the 100Mbit ports and Giga ports on 100Mbit link (rockets).
Testing from Rocket to RB2011 (5m FTP cable) ping 65000byte - 100% loss
Testing from RB912 to RB2011 (5m FTP cable) ping 65000byte - 0% loss 3ms
Colleagues replace RB2011 whit switch TP-Link 3210 and all running OK
Now the RB2011UAS is in my lab and testings starts
Reset to defaults , put IP on interface and cables
First test (first image):
RB2011UAS — Giga switch — Linux router
All test was OK ping from Linux to RB on 65500bytes is 2,7-3,0ms and 0% loss
Second test (second image):
RB2011UAS --100M-8port switch – Giga switch — Linux router
Same ping was 100%loss
PP---------------------------
Then try on 100M ports (port 6-10) has the same problem.
Strange in the whole situation is that the graph shows two-way traffic on RB2011 port and replays on Linux tcpdump, but there are 100% losses!
I tried with different versions of firmware and RouterOS, same results!
Is this hardware bug on 2011 series , because I have two RB2011 and two are held in the same way?
Sounds all like there is a problem with fragmentation. Why are you testing with 65000byte? In your 2nd image the dump looks like you are getting responses on the interface but the ping process didn’t get them. Are you sure that these responses are not filtered on that linux box ( iptables )?
My laptop (100M) — RB2011UAS local ping tests:
Pings whit 56710bytes is OK , but whit 56711bytes are 100%
Board is factory reset and set bridge whit IP
More testings , now pass thru capabilities
Linux --switch— 100M --(ether2) RB2011 (ether6)-- 100M — Laptop
ping 65500 bytes - 12ms , 0% loss
Linux --switch 1G --(ether2) RB2011 (ether6)-- 100M — Laptop
ping 65500 bytes - 100% loss
max packet size without (minimal) losses ~ 58000bytes
Linux --switch 1G --(ether2) RB2011 (ether5 gigabit port)-- 100M — Laptop
ping 65500 bytes - 100% loss
max packet size without (minimal) losses ~ 33000bytes
PP:
I guess that the problem in my previous post with ping to CPU (bridge IP) is similar because CPU us gigabit connected internally
Looks like you having problems once you use Gbit. These packets get fragmented anyway. So i suspect the fragmentation as you problem source. What is the MTU at your interfaces? are the Linux boxes starting to use jumbo frames when you using gigabit, so fragments are to big to be handled by the RB2011?
My guess is that is Rx/Tx buffer which gets full when you go from 1000Mbit → 100Mbit which causes dropped packets, but no counters are increasing under ‘/interface’ nor ‘/interface ethernet’ so it’s hard to tell.
There are no issues when you do 100Mbit → 100Mbit or 1000Mbit → 1000Mbit, but there you have a 1:1 relation in linerate.
I’m about to do a “well documented” lab and I hope I can take some time for that soon. And I will send my info to support@mikrotik.com when I have some results.
Are there anyone who have reported this to support?
Hello,
We have similar problems on RB2011LS-IN with 65000 packet size icmp ping
Linux → 1gbps → 1-6port on RB2011LS-IN → ok
Linux → 100mbps → 1-5port on RB2011LS-IN → 100% loss
Linux → 100mbps → 1-5port on RB2011LS-IN → 100% loss
Linux → 100mbps → 6-10port on RB2011LS-IN → 100% loss
I think its some hardware bug if mikrotik does not respond
I’m not quite sure if this is a bug. It may also be a limitation of the driver or hardware. There are other network devices like for example the Cisco ASA 5510, which aren’t able to handle 65k.
