TL-SG2008P LACP issue with Synology DS220+
This thread has been locked for further replies. You can start a new thread to share your ideas or ask questions.TL-SG2008P LACP issue with Synology DS220+
This thread has been locked for further replies. You can start a new thread to share your ideas or ask questions.
Hello,
New Omada user here, having an issue with TL-SG2008P and my Synology NAS with LACP.
My switch is managed by an OC200, LAG1 LACP enabled:
From Synology's end, LACP is enabled and both ports connected, link speed appears as 2000mbit (same results with static lag and balance xor)
however when doing file transfer to/from two other pcs which are connected directly to the switch, the total bandwidth appears to be 1gbit.
if i remove the lag/lacp configuration and use "adaptive load balancing" on the synology alone, the correct speed is achieved, so it looks like something on the switch is not being applied correctly with regard to LACP.
would appreciate any hints on this, seems like i'm missing something here
thanks!
- Copy Link
- Subscribe
- Bookmark
- Report Inappropriate Content
“in both scenarios, Layer2/3 LACP should be able to split the load equally and allow total bandwidth of 2Gbit from the NAS, is this not correct?”
Well, it is not correct. The process of selecting a port for any data flow uses a pure mathematical algorithm. The volume of the traffic does not play any role there. The key part of that algorithm is the calculation of those hash values. You have a choice of what that calculation is based on, usually MAC and/or IP addresses, but you will not know how exactly that calculation is done. To my knowledge, each company has its own proprietary hash calculation formula and keeps it secret. If there are just a couple traffic sources, there are just a couple of hash values and the chances that they match one and the some port are very high. If there are plenty of traffic sources, there are many hash values so the chances that they match different ports increase substantially. That usually results in a better load distribution, but not always. For example, if a certain source produces the majority of traffic, most of the traffic, of course, will be flowing by just one port. Only the other hand, even if there are only two traffic sources, you may get “lucky” and two different ports get matched. Like it or not, that’s how it works in those kind of LAGs and it is not just TP-Link specific.
- Copy Link
- Report Inappropriate Content
That just works as designed. The Synology’s Dynamic Link Aggregation is basically Balance XOR with the benefits of LACP. I’m pretty sure that the Synology’s Balance XOR is based on a hash derived from the source and destination MAC addresses so only one and the same Synology NIC is used when transmitting data to any given destination. The similar situation is on the switch side, but TL-SG2008 offers some more hash calculation algorithms. I’m not sure whether Omada lets you select them and, even if it does, the max file transfer speed between any two devices will be limited by the speed of the link which is actually 1Gb. If Omada shows the link speed as 2000Mb, it is totally unfair to make users believe in it.
- Copy Link
- Report Inappropriate Content
As @KJK mentioned ... this is as designed and works best for multiple clients accessing the server vs single client transfer speed increases, but what you might want to look at is Synology recently released support for SMB3 Multichannel which doesn't use LAG\LACP but will combine multiple individual connections automatically to use both connection bandwidth (even wired and wireless). Here's more info on that - https://kb.synology.com/en-us/DSM/tutorial/smb3_multichannel_link_aggregation - it's supported on Windows, Mac and most *nix and QNAP has it in beta release right now - most reports are that this helps with exactly what you were hoping to get from LACP\LAG.
Recent reddit thread on it for Synology: https://www.reddit.com/r/synology/comments/131byvt/smb3_multichannel/ for more experiences.
I've been looking forward to this myself. :)
- Copy Link
- Report Inappropriate Content
I understand that traffic between PC1 and NAS will be limited to 1Gbit, the point is that under no LACP or Static LAG settings am i able to break past the 1Gbit barrier (from NAS) even when hitting it from multiple PCs at the same time,
I should be able to fully saturate 2Gbit from the NAS towards PC1, PC2 and Laptop1, while both PCs are linked via their own 1Gbit ports, and Laptop is on an EAP225 which has its' own 1Gbit uplink.
In that scenario, under LACP/LAG, traffic is split across both NAS' NICs, however not fully saturating them.
if i turn off LACP and use the built in "adaptive load balancing", i'm getting the full bandwidth (2Gbit)
I have tried to change the hashing algorithm but the results remain the same under src-mac / dst-mac / src-dst-mac / src-ip / dst-ip / src-dst-ip
in the switch CLI it seems everything is configured properly, switch is passive and NAS is active, all looks good and honestly not sure what i'm missing here
Regarding the SMB3 option, i'm aware of it, but if i give up on trying to fix LACP, i might aswell just use synology's adaptive load balancing which gets me the same results across all network traffic, not just SMB..
TL-SG2008P#show etherchannel detail
Group: 1
----------
Group state = L2
Ports: 2 MaxPorts = 16
Protocol: LACP
Ports in the group:
-------------------
Flags: S - Device is sending Slow LACPDUs F - Device is sending fast LACPDUs.
A - Device is in active mode. P - Device is in passive mode.
Local information:
LACP port Admin Oper Port Port
Port Flags State Priority Key Key Number State
Gi1/0/6 SP Up 32768 0x1 0xe36 0x6 0x3c
Gi1/0/7 SP Up 32768 0x1 0xe36 0x7 0x3c
Partner's information:
LACP port Oper Port Port
Port Flags Priority Dev ID Key Number State
Gi1/0/6 FA 255 9009.d006.890e 0x9 0x1 0x3f
Gi1/0/7 FA 255 9009.d006.890e 0x9 0x2 0x3f
- Copy Link
- Report Inappropriate Content
Dynamic Link Aggregation may or may not split network traffic between NICs. It all depends on the calculated hashes. I have two Synology NASes configured with Dynamic Link Aggregation. I’ve done it for redundancy, but out of curiosity I have run speed tests for each NAS. I have used perf3 to run those tests. I used different workstations and servers with each NAS and ran two iperf3 sessions simultaneously in each test. In some cases, iperf3 indicated transfer speed between 930Mb/s to 970Mb/s in each session. That meant that the calculated hashes resulted in utilizing both NICs of the NAS. However in some cases, iperf3 indicated transfer speed about 430Mb/s in each session. That meant that the calculated hashes resulted in utilizing only one NIC of the NAS.
Adaptive Load Balancing does not use hashes when selecting NICs. The name indicates that the NAS actively measures the performance of each NIC while making NIC selection decisions. Unfortuately Synology does not provide any information on how it is actually done and whether it does not produce any undesireble effects. Switches do not offer this kind of load balancing.
- Copy Link
- Report Inappropriate Content
@KJK Thanks, I did some further testing like yourself and found that none of the 6 hashing algorithms provide 2gbit when going to wired clients, however when piling on wireless clients i'm getting 1.5~1.9gbit output.
Feels like something is wrong with the way it works "by design" but I'll just leave it for now and go back to adaptive load balancing, it just works 🤷♂️
- Copy Link
- Report Inappropriate Content
Give the switch a higher priority than the NAS and, with a little bit of luck, you will get different results. If a LAG is under LACP, the device with the higher priority (the lower number the higher priority) is responsible for selecting the active interfaces.
- Copy Link
- Report Inappropriate Content
@KJK That makes sense, previous output shows NAS end with priority of 255 while switch is defaulting to 32768
i've found this how-to https://www.tp-link.com/us/support/faq/991/ but that applies to switch standalone mode, for the life of me i can't find that in the omada controller, is it possible those settings are simply missing in controller mode?
- Copy Link
- Report Inappropriate Content
I have edited my previous post. It is actually the LACP system priority, not the LACP port priority. I don’t use Omada with my switches, so I can’t help you with it. However, I’ve noticed that Omada allows you to use CLI on certain products. If your switch is one of them, you may give it a try. The CLI command would be lacp system-priority pri in the CLI config mode.
- Copy Link
- Report Inappropriate Content
@KJK got it, my switch is not yet supported by device cli configuration, I'll have to test this after a firmware upgrade.
thanks again!
- Copy Link
- Report Inappropriate Content
Updating since my switch got CLI support this week finally,
i have set lacp system-priority 100, but no luck.
after that i have changed the ports which are connected to the NAS, updated the LAG config with the new ports, and set both ports individually to lacp port-priority 100 and again no dice..
(after both commands i did a sh run and sh etherchannel detail to verify the config was successful)
tried switching between the different lacp algorithms but that changed nothing.
NAS to one wired and two wireless clients achieves 2gbit,
NAS to two different wired clients tops out at 1gbit
starting to think i hit some bug here..
- Copy Link
- Report Inappropriate Content
Information
Helpful: 0
Views: 2772
Replies: 16
Voters 0
No one has voted for it yet.