[going crazy] host no more reachable after a while...

[voodooFX]

Maybe here someone could help me because I'm running out of ideas...

I have the following 4 servers

- esxi01
- esxi02
- freenas01
- freenas02

All of them are connected to this switch: Mikrotik CRS226-24G-2S+RM

in short, the problem is that ALL the virtual machines (on both esxi hosts) are UNABLE to reach/ping the freenas01 after some time

- if I shut down and start up one VM, it will be able to reach freenas01 for 30-40min, then nothing
- if I clear the arp cache: the same
- ALL the physical host have no problems to reach freenas01
- ALL the VM and physical host have no problems to reach freenas02

The freenas hosts have the same configuration

- on the switch: port trunk
- on the host: Link aggregation - Load Balancing mode

Does someone sees any logic in this?

 

[realtomatoes]

seen that in the lab
https://kb.vmware.com/selfservice/m...nguage=en_US&cmd=displayKC&externalId=1004109
i generally stopped using lacp.

 

[ttabbal]

My guess is some kind of compatibility thing with LACP. Some OS only support one mode of LACP, and some switches don't implement it properly in the first place. FreeNAS is one of the more picky ones, from what I can tell.

I had a fun one when I was playing with it. Local stuff worked fine, but the FreeNAS was unable to reach the internet. Everything else could pass through the router, but with LACP on, FreeNAS couldn't. Never did figure that one out. I realized the cost of 10GbE used gear had come down a lot since I last looked, so I picked up an LB4M and a pair of ConnectX2 cards and didn't look back. Most of my network is 1Gb, with a few random 100Mb like Raspberry Pi media players. So this works well for me and I don't have to worry about any goofy networking issues caused by LACP. I suspect that with 10Gb prices dropping, even for new, LACP just doesn't get a lot of testing. So compatibility issues just aren't discovered, or prioritized very high when they are.

Even at minimum wage, I spent more in time trying to fix LACP than I did buying the 10Gb gear. I'm firmly in the "LACP considered harmful" camp now.