Seems I have strange issue which to be started after I’d upgraded to 6.48.
I have small home network containing CCR1009-7G-1C, 2x CRS210-8G-2S+ and CRS 328P-4S+.
The CCR is the main router and GW. Anyway - worth mentioning all links are 1Gb or 10Gb.
The CCR provides internet to all through one of the CRS210 ports. (CCR is still the GW).
The CRS are in the role of edge switches and they are connected by optic over 10GB SFP+
My PC is connected on CRS210 and my NAS (and other servers from my home lab) are connected to CRS238.
My issue is that when I copy files from/to NAS my speed did not reach more than 300Mbps.
This is quite strange as the bandwidth of the network and the throughput of the arrays allows at least 600Mbps.
Does some of you guys have similar issue?
Any idea what maybe causes this issue?
I have even created a bonding ports to the NAS but this did not solve the problem.
If you downgrade to your previous release, is the issue gone ?
If so, open a bug-report.
This 6.48 release seems to cause a lot of issues on various architectures. I’m not updating my equipment for sure…
PS : I assume you are talking MBbps (MegaBYTES/sec) , not Mbps (=MegaBITS/sec)
I think before my systems was on 6.46.5, but since this all start with try to move to 7.1 beta… i moved to most updated stable firmware after the beta have failed.
Maybe worth downgrading back there.
The test shows that when my PC is on same switch with the NAS (CRS328) then I can achieve 112-114MBps which corresponds to 896-912Mbps.
So there should be something with the buffers or with the SFP+.
But those SFP modules was working well before…
In my previous post I specifically used Mb instead MB.
Windows/Linux shows in MB since Switches are in Mb where 8 bit=1 byte. In the network the speed is more correct to be calculated in bits as is “serialised” data transmission. Historical stuff, bits & bytes and so on - you know your 101.
What is not clear from your OP (at least to me) is the L2 and L3 topology. Since you wrote now that the PC can be connected to the same CRS like the NAS, I assume the PC and the NAS are in the same subnet and (V)LAN, so the traffic between the two is a pure L2 one (need not be routed by the CCR), please confirm.
If true, are the two CRS interconnected directly, or are both connected only to the CCR so the CCR has to bridge the traffic between them? Unlike the older CCR1009-8G-1S, the CCR1009-7G-1C doesn’t have any switch chip, so if bridging performance got affected by the upgrade, it can have this impact. What does /tool profile show on the CCR while the file transfer between the PC and the NAS is ongoing?
This is not a bridge port…
/interface bridge port
add bridge=bridge-LAN interface=ether3
add bridge=bridge-LAN interface=ether4
add bridge=bridge-LAN interface=ether5
add bridge=bridge-LAN interface=ether6
add bridge=bridge-LAN interface=ether7
add bridge=bridge-LAN interface=ether8
add bridge=bridge-LAN interface=sfp-sfpplus1
add bridge=bridge-LAN interface=sfpplus2 add bridge=bridge-LAN interface=vlan99 pvid=99
add bridge=bridge-LAN interface=ether1
add bridge=bridge-LAN interface=ether2
If VLAN99 is on the bridge and you have the bridge giving out DHCP,then why
/ip address
add address=192.168.1.2/24 interface=bridge-LAN network=192.168.1.0
add address=192.168.99.2/24 interface=vlan99 network=192.168.99.0 ???
The switches have static addresses while some machines and VMs not.
This is why the address is added to the VLAN interface.
But I saw something which you pointed: add bridge=bridge-LAN interface=vlan99 pvid=99
Seems this caused the traffic to go via CPU.
Please advise how to configure VLAN which to be available via all ports and have static address on this switch.
Maybe another bridge? But then how I’ll add the ports which are already added to another bridge?
What I was actually interested in was the other part of my post - whether the PC to NAS traffic is bridged via the CCR (or possibly even routed through it if the PC gets its IP from a DHCP server) or whether it can bypass it when the PC and the NAS are connected to different CRS.
@anav’s remark is also very relevant - it is not a good idea to attach both the tagged end of the /interface vlan and its untagged end to the same bridge. So just remove that red line from /interface bridge port, it won’t break anything. You don’t need to do anything special to permit VLAN 99 as tagged on all the member ports of the bridge-LAN except listing those ports in the tagged list on the row with vlan-ids=99 in /interface bridge vlan.
If the PC and the NAS discussed above are normally in different IP subnets, it means that the traffic between them must run through the CCR in order to be routed. If the two devices are normally in the same IP subnet, you could connect the switches to each other directly (and only one of them to the CCR), so that you wouldn’t waste the CCR’s CPU power on bridging.
And you still haven’t answered what does /tool profile show at the CCR when you test the transfer between the PC and the NAS, nor whether the only change between “all good” to “all slow” was the RouterOS upgrade on the CCR.
Thanks for reminding me Sandy.
I had checked this before. Yes they are in the same IP subnet.
Also I have defines specific routes which to avoid passing this traffic to CCR - it goes only via certain interface if is from the proper address.
I did not need to pass the VLANs by this switch until now.
Unfortunately I’ve found that when testing the speed drops immediately after I activated the VLAN filtering on the bridge port.
As expected the CPU hit 100% when transfer started.
I am just wandering what was the trick to avoid this traffic to pass via the CPU. There was some sort of specific configs.
But seems that according to this documents https://wiki.mikrotik.com/wiki/Manual:Interface/Bridge#Bridge_Hardware_Offloading
CRS210 was not supporting HW offloading with VLAN filtering. So may be necessary to ditch this switch.
CRS610 will be good swap but seems that also have issues with VLANS.
Maybe have to return temporary to my old CISCO or to put my spare CCR2004 (is not exactly switch but for short period will be ok).
Huh,
Sorry fellas!
I have felt victim on my own disinformation - i have forget to look the specs of CCR210-8G-2S+.
They do not support HW offload for VLANs and this is what was affecting the performance.
There a way this to be partially avoided with Switch chip configuration, but this will bring future issues as I have now do tests.
I’d just forget when I bought this devices and was thinking that if they have SFP+ it make them modern enough to crunch whatever I throw at them. Wishful thinking…
I have passed the difference in the configs between series 1XX,2XX vs 3XX during my MTCRE but have forgot it.
My tutor specially was pointing that on old devices some of the setups will require switch chips.
Anyway, thanks for the comments - they helped me to focus on the right point (and not to look in to buffer issues pointed for the more modern devices).
I think in few days will change it with CCR610 or CCR312. (still consider over the prices and necessities)
It’s not that bad actually - what the CRS1xx/2xx do not support is seamless integration of configuration of VLAN filtering in switch chip with the bridge configuration (this is only supported in CRS3xx), but you can configure the switch chip directly. Unless your intention is to use MSTP or some specific bonding modes in near future, you don’t need to replace the CRS2xx right now, it is enough to configure the VLAN handling on the switch chip. The right choice depends on the comparison of the price of your time to the price of the new switches.
Time is always priceless! You can earn money, but not time.
But if you get more knowledge, you can spend less time on the task
Experience is proven knowledge
