LAN setup from hell
Go to solution
Solved by SpaceGhostC2C,
So... I think I kind of sorted this out. It's been more of a blog than a thread, but in case anyone ever googles something similar, let's give it closure.
So, it kinda was the Switch in a way. It seems that proper 802.3ad LACP is feasible, but hard for it. And the implementation probably ever expected to be used between Cisco switches and some corners were cut or something. I'll try to break it down in parts:
- "Smartport" roles: only Windows (and possibly other Cisco switches, can't test) seem to interact successfully with "switch" ports. PFSense can to some extent, but not ideal. No Linux flavor could do it, although Ubuntu Server got a bit further (ping locally, no route to WAN), but I don't know how and could not reproduce it. "Desktop" (to a laptop), "Router" (to PFSense) and "Server" (to PC3) all worked well with the corresponding OSes. That takes care of most of Problem 2.
-
Link aggregation: my switch can't do LACP with any port role other than "switch", but I still got teams / bonds / laggs to work. First, I noticed I made a mistake in previous post: the Intel NIC in PC1 (Windows) was set to "SLA", not LACP, consistent with the Switch's Etherchannel being "static", making me think that probably LACP wasn't doing that well with Windows either (static Etherchannel does work, though). Intel's SLA is equivalent to LACP in my case, since I'm really not doing anything dynamically, everything is fixed both in hardware and logical terms, and both modes balance outgoing traffic across interfaces and receive incoming on any. I looked for similar static options in PFSense and CentOS: for PFSense, "LOADBALANCE" looked equivalent to me, with the benefit of specifying "no switch negotiation required" (which makes me think a simple unmanaged switch would suffice? Not sure). Similarly, I went with "balance-alb" (mode 6) in CentOS, with similar description and no switch support required. I removed all Etherchannels from the switch except the Windows one, and set the port roles individually as mentioned above. I now have a working bond in CentOS, a working lagg in PFSense, a working team in Windows, and all (reported) link speeds are 2.0Gbps
Oh, and tried rebooting PC3 while monitoring from PC1, and no experienced no downtime and no "possible flapping" message!
Now it's time to settle on a server OS for PC3, extend its bond to x4, complete the x2 team in PC2, and wait for the switch to catch fire so I can justify getting one from this century

Create an account or sign in to comment
You need to be a member in order to leave a comment
Create an account
Sign up for a new account in our community. It's easy!
Register a new accountSign in
Already have an account? Sign in here.
Sign In Now