F5F Stay Refreshed Power Users Networks Creating a LAN setup is incredibly challenging.

Creating a LAN setup is incredibly challenging.

Creating a LAN setup is incredibly challenging.

Pages (2): Previous 1 2
T
Tilzy
Junior Member
Find
38
10-09-2023, 11:38 AM
#11
So... I believe I've clarified things. It's mostly been a blog post rather than a thread, but let's give it a proper wrap-up. Essentially, it was about the Switch setup. Proper 802.3ad LACP seemed possible, but challenging to implement. It's likely that the intended approach was limited to Cisco devices and some compromises were made elsewhere. I'll try to explain it in sections: "Smartport" roles – only Windows (and possibly other Cisco switches, which I couldn't test) seemed to work well with "switch" ports. PFSense had some compatibility, though not ideal. No Linux options seemed viable, except for Ubuntu Server, which could ping locally but didn<|pad|>'s end reached (no WAN route). "Desktop" connections to a laptop, "Router" to PFSense, and "Server" to PC3 all functioned smoothly with their respective OSes. That mostly solved the main issue. Link aggregation was tricky – my switch only supported LACP on the "switch" port role, yet teams, bonds, and lagged properly. I realized earlier I'd misconfigured the Intel NIC on PC1 (Windows) as "SLA" instead of LACP, which aligned with the Switch's static Etherchannel. That suggested LACP might not have been handling Windows well either (static channels do work, though). Intel's SLA is similar to LACP in my case, since I wasn't doing dynamic negotiation and traffic balanced across interfaces. I checked PFSense and CentOS for static equivalents: PFSense offered "LOADBALANCE" which matched, with the note that it didn't require switch negotiation – implying a simple managed switch might work. CentOS had "balance-alb" (mode 6), which also worked without needing switch interaction. I dropped all Etherchannels except the Windows one and set port roles accordingly. Now CentOS has a working bond, PFSense a lagg, Windows a team, and all reported speeds are 2.0Gbps. I rebooted PC3 while watching from PC1, and there was no downtime or flapping messages. Now it's time to decide on a server OS for PC3, upgrade its bond to x4, extend the Windows team to x2, and wait for the switch to perform optimally so I can justify acquiring one from this era.
Reply
Reply
T
Tilzy
10-09-2023, 11:38 AM #11

So... I believe I've clarified things. It's mostly been a blog post rather than a thread, but let's give it a proper wrap-up. Essentially, it was about the Switch setup. Proper 802.3ad LACP seemed possible, but challenging to implement. It's likely that the intended approach was limited to Cisco devices and some compromises were made elsewhere. I'll try to explain it in sections: "Smartport" roles – only Windows (and possibly other Cisco switches, which I couldn't test) seemed to work well with "switch" ports. PFSense had some compatibility, though not ideal. No Linux options seemed viable, except for Ubuntu Server, which could ping locally but didn<|pad|>'s end reached (no WAN route). "Desktop" connections to a laptop, "Router" to PFSense, and "Server" to PC3 all functioned smoothly with their respective OSes. That mostly solved the main issue. Link aggregation was tricky – my switch only supported LACP on the "switch" port role, yet teams, bonds, and lagged properly. I realized earlier I'd misconfigured the Intel NIC on PC1 (Windows) as "SLA" instead of LACP, which aligned with the Switch's static Etherchannel. That suggested LACP might not have been handling Windows well either (static channels do work, though). Intel's SLA is similar to LACP in my case, since I wasn't doing dynamic negotiation and traffic balanced across interfaces. I checked PFSense and CentOS for static equivalents: PFSense offered "LOADBALANCE" which matched, with the note that it didn't require switch negotiation – implying a simple managed switch might work. CentOS had "balance-alb" (mode 6), which also worked without needing switch interaction. I dropped all Etherchannels except the Windows one and set port roles accordingly. Now CentOS has a working bond, PFSense a lagg, Windows a team, and all reported speeds are 2.0Gbps. I rebooted PC3 while watching from PC1, and there was no downtime or flapping messages. Now it's time to decide on a server OS for PC3, upgrade its bond to x4, extend the Windows team to x2, and wait for the switch to perform optimally so I can justify acquiring one from this era.

Reply
Reply
Pages (2): Previous 1 2