0

How to make sure my link on a host is not a part of a LACP (LAG)?

I'm trying to start booting from PXE but DHCP-requests don't reach DHCP-server most of time. I only can boot intermittently from one host of four. However, installed Linux OS can obtain an IP address. So I wonder if link aggregation is set on a switch, but network owner just provide me incorrect information saying that they have disabled aggregation

Макс Кринж
  • 13
  • 3

1 Answers1

0

For the LAG in general - you can't. Lots of vendor-specific LAG implementations are passive and do not emit anything, like a classic EtherChannel without PAgP or LACP.

For LACP (802.3ad) configured on the uplink switch in active mode do a Wireshark capture when link is up and look for LACP BPDUs with filter ether proto 0x8809. If you see anything it means that uplink expects this link to became a LAG member.

The above won't work if uplink is in LACP passive mode or any old-school pre-configured modes like PAgP that expect that other side is configured symmetrically.

Peter Zhabin
  • 2,696
  • 9
  • 10
  • I tried to setup bond-mode 802.3ad and it worked. I've seen in dump 0x8809 frames from server's interface to 01-80-C2-00-00-02 But if set active-backup mode it works too, without sending that frames. I still cannot figure out if the fact that bond works in 802.3ad mode means that LACP is configured on the switch side. Or it doesn't – Макс Кринж May 09 '23 at 18:04
  • Do you see LACP BPDUs from the uplink when bond mode is set to 802.3ad? If your bond works, please provide the output of `cat /proc/net/bonding/bond0` when it is up. – Peter Zhabin May 09 '23 at 18:45
  • It reads: `Bonding Mode: IEEE 802.3ad Dynamic link aggregation` `Transmit Hash Policy: layer3+4 (1)` `MII Status: up` `MII Pulling Interval (ms): 100` `Up Delay (ms): 0` `Down Delay (ms): 0` `Peer Notification Delay (ms): 0` `802.3ad info` `LACP rate:‘slow` `Min links: 0` `Aggregator selection policy (ad_select): stable` `System priority: 65535` `System MAC-address: 60:b3:11:21:85:40` `Active Aggregator Info:` `Aggregator ID: 1` `Number of ports: 1` `Actor Key: 21` `Partner Key: 1` `Partner Mac Address: 00 00:00:00:00 00` – Макс Кринж May 09 '23 at 18:59
  • If this a complete output? If it is it means that your bond interface has no slaves linked to it and is defunct. If there were some slaves listed that you didn't include and bond actually works, please post complete output. – Peter Zhabin May 09 '23 at 19:12
  • When bond mode is set to 802.3ad I only see this in pcap: 60:b3:11:21:85:40 > 01:80:c2:00:00:02, ethertype Slow Protocols (Ox8809), length 124: LaCPv1, length 110 – Макс Кринж May 09 '23 at 19:14
  • full /proc/net/bonding/bond0: https://pastebin.pl/view/e86b2a08 It has slaves, I just omitted them – Макс Кринж May 09 '23 at 19:20
  • All of your interfaces have partner as churned, so the LACP negotiation with peer has failed (which is expected as you see no LACP BPDU coming from the switch). So LACP is definitively not configured upstream in any form. For other forms of LAG see my original response, there's no way you can tell if any form of silent EtherChannel (like active-backup mode you have tried your side) is configured upstream. You can collect `tcpdump` on all interfaces while trying to send your DHCP requests on one of them to see if response comes on a different link, that will indicate a config issue upstream.. – Peter Zhabin May 09 '23 at 19:44
  • Thank you, very helpful! – Макс Кринж May 09 '23 at 19:54