Hello,
I’m using RB260GS with 1.12 firmware.
I disabled flow control on all ports but I’m still seeing TX pauses frame on single ethernet port.
Why the switch is sending out TX pauses frames with flow control disabled ?
I’m attaching TX & RX screenshot.
I also have problems, with pause frames bringing down the network even with flow control on the switch disabled. One customer has an old broken D-Link router which hangs after about 10-15 minuts uptime and then starts spewing pause frames like mad (flow control can’t be disabled on that router). This causes other customers connected to the same switch to lose connection (link still up but no PPPoE traffic can flow), the switch blinks all LEDs and its IP becomes unreachable too. Even ACL configured to pass only PPPoE frames (8863/8864 unicast in both directions, and 8863 broadcast from customers) doesn’t help, pause frames seem to bypass it. The RB250GS (latest firmware - 1.13) doesn’t crash, everything returns to normal as soon as that broken router is turned off. I had the same issue with a cheap non-managed TP-Link 5-port switch, and hoped to solve it with the RB250GS. Or do I need a RB750 for this kind of bridge filtering (block pause frames)? Don’t need gigabit, it’s all behind a LocoM5 anyway (which fortunately doesn’t pass these evil pause frames further upstram, as that could bring down the rest of much larger network).
I am also very interested.
how i can protect my network against pause frame flood? I had CRS as central switch
If this can be of any help to the MikroTik team, I have attached the output I get from running ethtool on a Linux system connected to the switch with the “flow control” option disabled and enabled.
As far as I can tell, the only difference when it’s enabled is that the switch advertises support for half-duplex link modes in addition to full-duplex ones.
This is still an issue in SwOS 1.16.
This is still an issue in SwOS 2.9!!! RB260GSP
Do you scoff?
I only discovered this after I bought 40 units and figured this out.
I know this is an old thread but I was able to correct the pause problems on my RB260GSP’s easily and wanted to share.
I was consistently experiencing pauses for 90% of 64k packets, so I would have 3mil packets with 2.6mil pauses on my VLAN Trunk ports. I found the RBs on the other end of those Trunk Ports had a default L2 MTU of 1514. After changing it to 1522 which includes the VLAN header, I have no ZERO pauses.
I hope this helps others.
thanks for sharing
Welcome, after 48h I do have some pauses but far less than I had originally. Just not sure it makes sense, so hoping someone else with more knowledge can suggest if and how this helps?
Just an update, I recently changed the MTU to 1600 and now ALL pauses have disappeared, network is much faster and I have far fewer 64k packets overall. This L2 MTU matches the value calculated for the cap interfaces by capsman, which includes VLANs.
nice to know that
@Checito do you know why increasing the MTU to 1600 would reduce the number of 64 byte packets by great amounts (like approx 100x less)?
which reference are that Routerboards ??
I have a CRS312-4C+8XG running in SWOS 2.13 and see similar issues.
I see RX and TX Pause frames on some switch ports although I have disabled Pauseframes on the switch.
As you said you have set MTU to 1600, this is was on the Application/Host side or on the Switch side? Because on my Switch I can’t set a MTU size.
Thanks.
The MTU changes I discussed were on my router not the switch, I hope that helps!
Hi
that did not work for me.. i have OLT with 10G uplink sfp+ port MTU 1500 and max MTU transmission 1598 , link passes trough a CRS in switchOS mode with same vlan on crs317 ports and then connect direct to CCR1036 SFP+ port.. on CCR1036 port we have MTU 1500 and L2MTU 1598 size..
On mikrotik ccr1035 side the sfp+ port has Flow control ON on both tx and rx..
On the OLT side the flow control also enabled on the uplink 10G sfp+ port
On the CRS switch on both ports connected to OLT and CCR.. flow control disabled , max speed we can get on wifi6 devices is 250Mbps.., after we enable on both ports from CRS switch the flow control.. we can get on the next speedtest 700Mbps on wifi6 devices connected on wifi..
it works great for a couple of hours until we start seeing RX-PAUSES with hundreds of thousands in the CRS stats logs… and after few hours.. we seek internet degradation… wifi6 speeds will not go over 150mbps , after 1.5gbps traffic on the 10G uplink por from OLT → CRS 10g port —> ccr1036 internet uplink sfp+ port… we start seeing alot of LINK PAUSE message on the CRS port..
so i am wondering how this can be fixed? if it can be fixed at all ?
i have tried changing the interface-queues from only-hardware-queue to multi-queue-ethernet-default.. but same issue happens after a few hours..