Does anybody have any insight on what Mikrotik is doing with MLAG? It appears that MLAG has been broken since 7.7 and looking at the changelog in 7.7 I see a few bridge changes:
*) bridge - added support for static MDB entries;
*) bridge - disallow port-controller while the bridge has MSTP enabled;
*) bridge - fixed “edge=yes” setting for MSTP;
*) bridge - fixed MSTP compatibility with STP;
*) bridge - fixed R/M/STP bridge identifier on protocol-mode change;
*) bridge - fixed RSTP BCP with bridged PPP interfaces;
*) bridge - fixed STP blocking state on port-controller;
*) bridge - fixed host moving with fast-path;
*) bridge - fixed incorrect root port blocking for MSTP;
*) bridge - fixed master port conversion;
*) bridge - fixed mst-override port priority for MSTP;
*) bridge - fixed port priority for STP and RSTP;
*) bridge - improved port-controller system stability;
*) bridge - improved system stability when using MSTP and many VLAN mappings;
*) bridge - removed “age” monitoring property from the host table;
Not saying that those changes are specifically what broke MLAG, but it’s probably related. MLAG was perhaps a little pre-maturely released, but it’s now completely broken.
I understand that sometimes things get broken while moving forward, but anything that gets messed up along the way really should be fixed pretty quickly. We’re coming up on 6 months that this has been broken. I need to expand some networks and those networks need some sort of redundant access-layer solution, and right now, Mikrotik is apparently broken with no fix in sight.
We also found that MLAG on ROS versions higher than 7.6 (so 7.7, 7.8 and 7.9) does NOT work, especially if you have two pairs of switches in MLAG (4 devices)
We also have 2 device setups with ROS 7.9 and MLAG works fine…
Interesting…I hadn’t seen your thread otherwise I would have piled on in there.
In my case I have a pair of CRS326-24S+ switches as a “core” with a pair of CRS328-24p switches below.
When everything is on 7.6, the MLAG on the “core” layer works great. Trying to enable MLAG on the CRS328 “access” pair just creates a big spanning tree loop. Similar MLAG config as with the CRS326’s above, but the 328’s fail and create a loop.
Upgrading the CRS326 pair beyond 7.6 causes instability. Spannihg-tree blocking, MAC shifting, it’s a mess.
I have another pair of CRS326-24G switches on an additional site for that customer and they experience the same issue moving past 7.6.
It’s unusable. I’ll take a look at your thread more in depth when I get some time tomorrow.
Hi, Mikrotik Support has reproduced and confirmed the bug:
Hello,
Thank you for the report!
We have managed to reproduce the issue locally in our labs and look forward to fixing it on upcoming RouterOS versions, unfortunately, I cannot provide a release date now.
Unfortunately, I cannot suggest any known workarounds.
Best regards,
Edgars P.
Please, tell me - is the MLAG working on 7.10?
I have issues with configuring the Mlag, not appear in active port in the following command:
Interface/bonding/monitor “name of the bonding”
Mikrotik MLAG has never worked properly. Anyone who says otherwise has never used a REAL working MLAG setup. The entire Mikrotik approach to MLAG is half-baked. Ive been waiting over a year for a fix and they could care less.. I have a pair of nice paper weights sitting here.. Completely worthless equipment.
Hi! Are you refering to the issue that the interfaces after a switch reboot starts sending traffic through only one switch? I’m trying to make 2 CRS326 24S+ 2Q working at the same time balancing the traffic of my network but is impossible. I have to change the mode of the bondings in the switches after a reboot and revert the changes. There’s no way to keep that working fine unless you use balance-rr but the interface performance is so much poor.
Hello,
Thank you for the report!
We have managed to reproduce the issue locally in our labs and look forward to fixing it on upcoming RouterOS versions, unfortunately, I cannot provide a release date now.
Unfortunately, I cannot suggest any known workarounds.
Best regards,
Edgars P.
I’m referring to the fact that they (mikrotik) have no idea what they are doing. Thr lacp system id changes when devices are rebooted. Causing disruption. The spanning tree bridge id changes when devices are rebooted. Causing disruption. It doest work with igmp snooping. It’s a poor attempt.
Since MLAG doesn’t work properly, how would you rather setup a redundant network, using 2x CRS326-24+2Q switches, connected to a RB301?
I have 5x Proxmox servers, setup with LACP (802.3ad) over 2x SFP+ ports. And each SFP+ port is connected to a CRS3262-24+2Q switch.
EDIT:
i just gave you alternatives RIGHT above your post
if you are on a budget … fs.com S5850 or N5860. i would recommend the N
definitely heavier on the price tag than mikrotik but quite solid switches
if budget is no concern you could go with juniper or extreme.
if you REALLY do not care about money - cisco nexus 9k, but licensing is an utter shitshow over there
You are throwing your money away on mikrotik if you intend to mlag. Plain and simple. 7x more and working is a better deal than 1/7 cheaper and doesn’t work at all.
Anyone living with a mikrotik mlag setup had never used a real good mlag setup from a number of other vendors.
As one of the people who heavily pushed Mikrotik to add MLAG and tested it from Day 1, I completely agree. They really missed the mark on this feature.
I wish they had brought a couple of cheap Extreme Summits from eBay and set them up in a MLAG to see how to do it properly before they implemented it.
