using RB2011L with latest update available to RouterOS. I’ve tried bandwith limits in past few days but always with the effect of affecting other communication too. My setup is RB2011L sitting in DMZ of Zyxel ADSL modem. To RB2011L there are connected two users to ether2 and ether4. To ether5 I do have connected smc switch with other users. Now, I’d like to limit bandwidth available to ether2/4 to just 512kbps up/down. When I do that my communication on ether5 is affected too. The not so funny thing is that more I limit ether2/4 more I’m limited on ether5. I’m setting bandwidth in web UI. In interfaces I select interface from the table and then set preferred bandwith limit tx/rx from the drop-down list.
Why I’m limiting to this silly speed is that for now I do have only 2Mbit down/256kbps up link and I need to preserve some bandwidth to ether5 users.
The behaviour of problem is as follows:
without limit, let’s say I ping RB2011 from machine connected to ether5 in 0.xxx ms. I ping Zyxel in 1.xx ms
with limit set on ether2 to 512kbps tx/rx my pings to RB and Zyxel suddenly increase to 100-200ms
with limit set on ether2 to 64kbps tx/rx (to test) my pings to RB ar on 1000-2000ms and to Zyxel 1500-3000ms
That’s while I’m downloading something on machine connected to ether2 and while users connected to ether5 are basically idle and not downloading anything. I’ve also verified on RB side and the CPU is loaded 1-3% so this is not caused by CPU limitation.
The problem is that such heavily affected latency also affects tcp behaviour and web stops working well…
Any idea how to solve this or workaround how to use this feature?
OK, I've just looked and my /export contains some private info I'd like to keep secret (dhcp server leases for example). Obfuscating this would keep some time. It would help to know what menu sub-trees are most important to export.
Otherwise I've tried hard to include any relevant information to allow MikroTik staff to attempt to duplicate the issue and just reply with their findings even if that may be "we're not able to duplicate your issue, post more info..."
Without seeing the config export it is hard to guess, but have you used ****
/tool profile
to see at which tasks your CPU spends most time and how much in absolute figures? The thing is that it may simply be so busy handling the traffic shaping that it is unable to deal with the overall throughput.
If you have a look at the block diagram for the 2011 series, you’ll see that ports 1-5 are attached to the same switch chip. https://i.mt.lv/routerboard/files/RB2011iL-160620170215.png (I know this isn’t exactly your model but I think everything but the nand memory is the same).
Do you have the option of changing port 5 to another port on the gigabit switch, and then another port on the fast ethernet integrated switch and test the issue again each time? That could rule out a configuration issue and point towards a hardware issue.
By the way, the command ‘export hide-sensitive’ will automatically obscure passwords and such, but you might still need to obscure DHCP leases manually, if you consider them private.
