LACP not prioritizing desired ports
LACP not prioritizing desired ports
I’ve got two 10 GbE fiber links and one 1 GbE Cat6 link between switch A (TL-SG3428XMP 3.0, 3.0.2 Build 20221130 Rel.57838) and switch B (TL-SG3210XHP-M2 2.0, firmware firmware 2.0.0 Build 20220322 Rel.62255). I want to use both the fiber links and fall back to the Cat6, so I’ve configured exactly one LACP group (ID 2, chosen at random) on each switch, containing the fiber ports at the default priority of 32768 and the Cat6 port at 65000. Switch A is the active side.
Whenever switch B is rebooted, the 1 GbE link is what comes up. The only entry on the LAG Table page is the Cat6 port, and the LACP Config page only shows LAG2 on that port (on both sides). If I unplug the Cat6 cable, both fiber ports become active (reflected on the LAG Table and LACP Config pages), and plugging the Cat6 cable back in later doesn’t change anything.
My understanding is that a lower number denotes a higher priority. Is there something I need to change in order to make it work correctly? I don’t see why it should be using the Cat6 link except as a fallback.
- Copy Link
- Subscribe
- Bookmark
- Report Inappropriate Content
Hi @shivjm
Thanks for posting in our business forum.
shivjm wrote
@Clive_A I can see that the LAG Table shows only the 1 GbE link whenever I restart switch B (see my screenshots from earlier) and I’ve verified with iperf3 that it is in fact using that link. I can also see the LAG Table update when I unplug the copper link, like in this screenshot from the passive side:
And I can see the difference in speed in iperf3 in that situation. Note that if I make the fiber links work like this and then plug the copper link back in, it doesn’t revert to using the copper link; it stays on the fiber links.
So I confirmed this with the senior engineers, don't LAG two different ports of link speed at the same time. It does not work. You can set it but it won't take effect. There is a potential that the MAC address table would go in error.
So, if you set a LAG group, set the same link speed ports in the same LAG group. Don't mix them up.
- Copy Link
- Report Inappropriate Content
The switches that do not allow me to activate LACP on LAGs with links of different speeds are actually CISCO small business switches. When I try to do that, and I have never tried that before your post, I get an error message complaining about it and the switch accepts only the 1G port as a LAG candidate. The 10G ports are droped from the candidate list.
- Copy Link
- Report Inappropriate Content
I might’ve misunderstand your posting, but here it goes.
A LAG with two links already offers redundancy and a third link is not really necessary. Also, you should not mix ports of different speeds in a single LAG and, actually, LACP should not allow you to do so. If you want to have that extra 1G copper link anyway, you can have it outside of the 10G LAG. That link together with the LAG will create a loop so you need to enable STP/RSTP to control the use of those links. Still, it would be a rather unusual configuration in my opinion.
- Copy Link
- Report Inappropriate Content
Thanks for the reply, @KJK. To be clear, I don’t want to combine the copper link with the fiber links, and I’m aware that two fiber links already offer redundancy. I just want to use the copper link as a fallback. My understanding from this part of the manual:
The port with higher priority in an LAG will be selected as the working port to forward data, and at most eight ports can work simultaneously.
Is that the copper link, whose port has a lower priority (65000), should only be used to forward data if the fiber ports (with a priority of 32768) are malfunctioning. Instead, the copper link seems to be used first, possibly because it comes earlier in the list of ports, and only when I physically disconnect it are the fiber links used instead.
In other words, unless I’ve misunderstood what’s happening, LACP seems to recognize that they have different priorities, but not the significance of those priorities.
- Copy Link
- Report Inappropriate Content
LACP manages the LAG only. If the copper link is not part of the LAG, the LACP settings do not apply to the copper link. You need to look at the STP/RSTP settings.
- Copy Link
- Report Inappropriate Content
I think we’re speaking at cross-purposes because I’m not using the right terminology; when you say LAG I understand that to mean a LAG negotiated by LACP (or a static LAG, but that doesn’t apply here), but I think you were talking about the LACP LAG. I have indeed put the copper and the fiber ports in the same LACP LAG because I want it to function as described in the manual:
One LACP LAG supports multiple member ports, but at most eight of them can work simultaneously, and the other member ports are backups. Using LACP protocol, the switches negotiate parameters and determine the working ports. When a working port
fails, the backup port with the highest priority will replace the faulty port and start to forward data.
I’m not trying to get LACP to negotiate a LAG containing links with mixed speeds. I want it to use the fiber ports, which have a higher priority, and re-negotiate if that fails so that the copper link is used instead.
- Copy Link
- Report Inappropriate Content
My switches do not allow me to activate LACP on a LAG that contains links of different speeds. If your switches allow that, I do not really know how it works.
Re. the LACP feature you quote. I understand it that it applies to a LAG that contains more than 8 links.
- Copy Link
- Report Inappropriate Content
Hi @shivjm
KJK was right partially about the statement on STP/RSTP.
Your description is not clear to me.
So my question is you have three links altogether, 3 are grouped in one LACP group. Is this what you are saying? If yes, that seems to be fine. STP/RSTP is not needed. If they are not in the same LACP, you probably should reconsider KJK's proposal on STP.
Also, you might wanna take a look at your log.
Your statement on the priority is correct. Smaller numbers mean higher priority. In the following discussion, please paste with screenshots to illustrate your issues.
- Copy Link
- Report Inappropriate Content
Thanks for the clarifications, @Clive_A. Yes, they’re in one LACP group. Here’s switch A (the active side), where port 9 is 1 GbE and ports 27 & 28 are 10 GbE:
Here’s switch B (passive), where port 1 is 1 GbE and 9 & 10 are 10 GbE:
Is there anything specific I should look for in the logs? I see ‘Added new LACP Link Aggregation Group 2 by console’ and later these messages (I’ve preserved the UI’s reverse chronological order):
Te1/0/10 changed state to up.
Te1/0/9 changed state to up.
Tw1/0/1 changed state to up.
- Copy Link
- Report Inappropriate Content
It is not from TP-Link, but this link may be helpful :)
https://www.cisco.com/assets/sol/sb/Switches_Emulators_v2_3_5_xx/help/250/index.html#page/tesla_250_olh/lacp_conf.html
- Copy Link
- Report Inappropriate Content
Hi @shivjm
Thanks for posting in our business forum.
What is the result of show lacp internal and show etherchannel? Screenshot?
- Copy Link
- Report Inappropriate Content
- Copy Link
- Report Inappropriate Content
Information
Helpful: 0
Views: 460
Replies: 17
Voters 0
No one has voted for it yet.