I have purchased two of these CRS switches for a project and wound up needing one in my data center to aggregate 10-12 servers…
I am trying to set it up like this
eth1-2 LACP VLAN trunk to CCR1036 router (vlan 88-93)
eth3-6 LACP VLAN trunk to top of rack cisco switch (vlan 88-93)
eth 8-24 access VLAN 93 (untagged)
I’m usually pretty good with LAN switches, but this thing has me scratching my head - is there some documentation or a config example to make this work? Thanks in adavnce…
Lacp and vlans don’t work correctly. We are waiting for them to fix it and then release more documentation and examples.
Sent from my SCH-I545 using Tapatalk
so would it be possible to ditch the LACP and just use e1 and e3 as trunks and the rest as untagged access ports like above? I read that if I use a bridge the CPU will kill my throughput, not sure how to configure the switch to do this?
Honestly I have no idea. I’m still trying to make these things work myself.
Sent from my SCH-I545 using Tapatalk
Yet another case of them putting out potentially good hardware without having the software to back it up.
STOP MAKING HARDWARE… FIX THE ROS. GET IT STABLE.
uggggg… this is soooo frustrating.
Hire another company to develop the software since you can’t seem to get it done right.
I tend to have to agree with this one sadly.
MT, you guys need to seriously invest more time in ROS and not in hardware. You won’t be able to please every single person with your device versions. But you can at least try and satisfy your current users.
I think releasing hardware, without the correct software to back it up, is absolutely stupid.
The CRS is a perfect example of it!
The unit is great, but the missing/broken features are a serious problem, and should not have been allowed to pass through. At the very least they should be addressed IMMEDIATELY. So far you haven’t even acknowledged the issues, or given an ETA.
These issues are very “basic” functionality of a managed switch. And right now they are a disaster. This would be like ROS not being capable of using the NAT function. It’s a complete waste for the functionality.
PLEASE ADDRESS THESE ISSUES!
it’s working ok, sort of, as just a straight layer 2 with eth1 plugged into and untagged port and e2, 3, 4 slaved, eth5 on another untagged port and e6,7,8 slaved… really aggravating not to be able to use a switch as a switch…
Yep. Hopefully it will be fixed soon.
Sent from my SCH-I545 using Tapatalk
I am having a few CRS125 in production. The only ROS which kind of works for me is 6.6 - anything else causing issues with speed on switched ports (10Mbit/s max) and/or sometimes weird hanging of either ports for (10-20 seconds) ip phones udp connection blend in/out. Just to keep it up and not having to replace the hardware I currently dropped all switching features and set it up as layer2 (bridge, vlans - uh CPU!).
Maybe it is worthwhile to mention, that while following the wiki how to configure CRS for vlans, I had not noticed any issues and worked as expected - it looks like ROS is working ok on its own / switching works ok on its own. Now mixing those together (as Mikrotik mentions it should be possible) just does not work. As I do also have a ‘mixed’ configuration - ie the vlan trunk is always redundant, therefore i require to bridge at least 2 ether ports and run rstp on it, the rest should be switched ports → no go.
Regarding CCR1036 - had a go in conjunction with a stacked Cisco 3750 lacp etherchannels - cisco reporting mismatched configuration on port channels after 40-50 seconds and shut down interfaces. Gave up after a couple of hours stinking around. - forgot to mention, using RB750G’s with same configuration and ROS 6.10 actually worked as expected 
In short - it seems semi complex setups for backbone utilization on CRS125 or CCR series (bridge, lacp, vlan , switch) seems currently not to work.
I believe Mikrotik has serious work to do to locate and fix all those bugs - being optimistic - give them another 4-6 month. I dearly hope they proof me wrong though.
EDIT:
missed - Bonding interfaces from CRS125 to other MT’s works fine with ROS 6.10 - Bonding interfaces to little Cisco SG200-26p works fine too. Did not test potential speed issues as CPU was simply 100% occupied with simple copy process - CRS125 are NOT equipped with powerful enough CPU to utilize bonded interface configuration and throughput speed actually drops compared to single port connection.
Guess I should have searched more diligently to find this post. I have almost the same exact issue. I can create a single trunk port, however, I am going to need at least 3 more trunk ports on the same device.
We provide wireless high speed data connections to rural businesses, and have several large towers that are distribution points to several smaller towers each. We are looking at putting CRS125’s or larger at each main distribution tower, and place another MikroTik of some sort at each smaller tower that supplies connection to several customers. We need the ability to propagate VLANs out to the smaller towers by setting up multiple trunk ports for each connection off the main distribution towers. As of now, all I have been able to do is trunk the CRS125 back to the main shop that feeds the distribution towers. Any further info is GREATLY appreciated, but it looks like, as others have stated, a bug in the RouterOS software.
Thanks!
Jay
We only just deployed these in a few locations over the past weeks and today we have just removed the last CRS125 from production replacing them with the hardware they were originally supposed to replace!
Unfortunately the problems we faced were too many to continue with in production. Until things like no isolation between vlans, speeds down as low as 19Kb/s between two Gig interfaces on the same switch are resolved we can’t use what looks like it could be a really great product. I’m going to have to call our vendor to discuss RMA options.
I really hope it’s less than that! I’m an huge advocate of Mikrotik within our company but problems like this don’t support my purchases. I look forward to the next release where I hope these issues are resolved or acceptable work-arounds can be found in the meantime. 6.10 looked to be a huge improvement in testing but a lot of the problems still remain.