NETGEAR is aware of a growing number of phone and online scams. To learn how to stay safe click here.
Forum Discussion
M4300 LACP/LAGG only works when incorrectly configured
I have video proof this time, although I'm not sure if I can post it.
I had a misconfigured M4300 where one of the hosts had a VHID that overlapped the LACP ports. The LAGG functioned correctly...
I have video proof this time, although I'm not sure if I can post it.
If you have a google account, you can post the video in YouTube. Then, post the YouTube video link here.
Having cleaned out everything, can tech support provide a working LACP example script?
I think NETGEAR Tech Support people are not able to monitor the NETGEAR community from time-to-time since they are more focused on the support tickets opened by NETGEAR clients. It would be best if you open a chat or online support ticket with NETGEAR Support at anytime with regard to your concern. Let us know how it goes.
Regards,
DaneA
NETGEAR Community Team
Well, the OP isn't able to provide any indication (links, documentation, whatever, ...) that these 40G cards or the host should be workable on x10GbE links, and wht configuration is required. In my understanding - unless there is something specifically configured on the card and/or the related host - with all due respect, any support attempt is a waste of time. We only get shown whats on the switch, nothing from the peer.
Well, hadn't realized that was what was required- but thanks for that hint!
The cards are Chelsio T580s.
https://www.chelsio.com/nic/40gigabit-ethernet-40gbe-adapters/
They are configured in 4x10gb mode, and quite capable of pushing 4x10gbE in LACP mode- in a Cisco switch.
https://github.com/freenas/freenas/blob/master/src/freenas/usr/local/bin/chelsio_adapter_config
The switch was reset to default configuration and ports were added to each of the LAGGs- 4x on each LAGG, and L3 hash was chosen.
The 'static' mode is disabled- cards will not work with that- and TrueNAS is set to LACP.
Running iperf3 forward with 8 threads shows 10gbE bandwidth. The crazy part- IF I uncheck the ports associated with the LAGG of the initiating machine (so, on 1/2/3/4 with machine 1), I can run iperf3 and see 40gb/s of transfer speed aggregate. This lasts for about 30 seconds, then the connection dies (since there are no ports associated with the LAGG anymore). Re-enabling all of the ports results in this running at 10gbE speeds again.
Even more crazy if I run iperf3 in reverse (using the -R) on machine 1, and uncheck the ports associated with machine 2, it'll run at 40gbe speeds until, again, the connection dies.
The only other time everything worked correctly was when an invalid port was assigned to the LAGG due to an ARP conflict. It's as if the switch simply can't accept a correctly working LACP setup.
Which brings me back to- if I've followed the directions to set up LACP correctly on the M4300, why does it only work during transitioning states?
So the LACP bonding of the four single interfaces is done on the card, or on the Linux Kernel bonding driver?
If it's Linux bonding, check e.g.
# cat /proc/net/bonding/bond0
I'll check that first thing Monday morning.
In the meantime, I have a video uploading. I had to redact portions of it.
https://www.youtube.com/watch?v=NuUF2Rkailo
I'm not a video guy so I have no idea how to enlarge portions of it to show you better, but at HD you should be able to watch the speed suddenly increase once I uncheck the ports associated with the LAG membership, and then decrease when I add the ports back in.
NETGEAR Academy
Boost your skills with the Netgear Academy - Get trained, certified and stay ahead with the latest Netgear technology!
Join Us!
