I'm planning on implementing MLAG with CRS510-8XS-2XQ-IN switches.
However, in preparation I've been reading forums a bit regarding MLAG, and most of them seem to be negative regarding it being unstable.
Some posts indicated that 7.19 is the safest I think, but that was a while ago.
It was also stated that more than 2 links per host was not "safe". I'm only going to use 2 links, but would be nice to know if there have been improvements.
I don't have the switches yet so I haven't been able to start testing myself.
So, where do we stand with MLAG now with 7.22/7.23? Is it usable and stable?
I have 7.23 in production on 3 MLAG pairs.
I wouldn't call it stable and do not advise using it. I will say it is functioning...But I do have some random oddities.
For example, I had one MLAG pair fail without any error messages in the logs. How did I know it failed? Well...Aside from the weird traffic reflection that was confusing the site router, the two nodes couldn't ping each other on a shared VLAN between the two. A reboot of (both) switches in the MLAG pair "fixed" it. I haven't had any troubles since, but that was just 3-4 days ago.
I had another pair fail in such a weird way that I can't even explain. Just flakey behavior. One example, I just just disabled an unused physical Ethernet port (interface ethernet set x disable=yes) and the whole switch locked up and took out the entire network. All traffic stopped forwarding on both switches in the pair and the switch wouldn't respond to IP OR the console port. I unplugged the power, waited a few, and powered it back up. That sufficiently screwed up the world though. I had to take down the VM farm and shut down the whole environment and bring it back up to fully recover. Again...nothing in the logs to indicate a problem.
For another MLAG pair running 7.22, (the "core switches" in this case), I setup CAPsMAN on core01. When the CAP connected to the CAPsMAN, I started getting the dreaded fdb update errors. IN this case I was able to reboot the core01 switch and let it link back up with it's peer. When it came back up, the CAP reassociated and there were no other fdb errors in the logs. It lasted about a week before another meltdown (fdb errors again). It's on 7.23 now...So far no errors, but it's only been up a week or so.
Use with caution. Sirbryan seems to have more success with 7.19.3 (I think) in his environment.
Hmm, well that's concerning. I will by running DRBD over it, so stuff like you're describing would be bad.
Moving to MLAG is not critical for me, but it would be very nice for the additional redundancy and ability to take down a core switch temporarily for maintenance.
Would you say it could be better to wait with implementation for me then? Do you know if MLAG stability is hi/low/non-existent on the RouterOS roadmap?