LAGG problem
-
Hello all,
Not knowing if I should post on redmine for a potential bug, I submit it here.
On my DIY pfsense firewall, I have a bridge between a 10Gb socket (on a PCIe card with two SFP+ sockets, the first being the WAN), 2 1Gb sockets on x11ssl-cf motherboard and 4 2.5Gb ports on a single PCIe card.
I created a LAGG with the two 1GB and one 2.5GB for my synology RS820+ NAS (itself well configured in LACP on the three ports), LAGG that I added to the bridge.
if I do a speedtest of the internet connection (freebox delta 10GB downstream connected to the SFP+ port) from the NAS, I saturate the LAGG at 2.83 GB.
if I do a speedtest from my PC I have 1.5GB downstream (PC connected to the 10GB pfsense port with a card in the PC in 5GB PCIe.
if I do an iperf3 between the NAS and the PC, I do not get more than 1GB (950 MB to be exact).
I added a 10 GB SFP+ card to my NAS. If I do an iperf3 again, this time I saturate the connection between the PC and the NAS with the maximum possible flow rate.My question is: is it possible that there is a bug? it seems that the flow rate goes well between WAN and a LAN or WAN and LAGG, but on the other hand the flow rate is restricted between LAGG and a LAN, to a single port of the LAGG.
Should I create a request on redmine?
Thanks for your opinions
-
@PiAxel said in LAGG problem:
flow rate is restricted between LAGG and a LAN, to a single port of the LAGG.
And that exactly how it would work - a lagg is 2 x 1ge ports not 1+1=2ge
You can load share sure - but if same mac talking to same mac or same IP talking to same IP, etc. all sessions would go over the same physical port in the lagg.
If you want device x talking y over 1ge - then both of their connections should be over 1 gig.. No lagging or lacp is going to get you 2ge over 2 1ge interface. If your talking smb you could do say smb3 mutlichannel and leverage 2 physical connections for the smb xfer.
-
@johnpoz said in LAGG problem:
You can load share sure - but if same mac talking to same mac or same IP talking to same IP, etc. all sessions would go over the same physical port in the lagg.
This should be in bold.
-
Mmm, I would expect it to sometimes end up on the 2.5G link if you repeat the test though.
However bridged connections can behave oddly.
