Wired Intelligent Edge

Hello,

 

We are installing a new VSX stack based on Aruba 8320 VSX stack runnin OS-CX 10.02 We have issue right now where the LAG is not working with HP 5500 partner. First we wer using VSX MCLAG but while it was not working I fallback to one link on LAG on the one switch (non mclag). Here is the case:

 

We want LAG between the 8320 and 5500 and the link is ethernet 1Gig with tranceivers J8177D on the Aruba side (SPF+ to Eth 1G). Link is up but the lacp is blocked from aruba side.

 

Here is a snippet from the config:

 

Aruba 8230:
--------------------------------
interface lag 100
no shutdown
no routing
vlan access 1
lacp mode active

interface 1/1/25
description csw20_1_0_9
no shutdown
lag 100

HP 5500
--------------------------------
interface Bridge-Aggregation120
port link-type trunk
port trunk permit vlan all
link-aggregation mode dynamic

interface GigabitEthernet1/0/9
port link-mode bridge
description testlink_to_new_core2
loopback-detection enable
loopback-detection action semi-block
broadcast-suppression pps 3000
port link-aggregation group 120

 

This can be seen in the event log for lacpd on Aruba 8320:

 

2019-02-13:09:50:27.364301|lacpd|1311|LOG_WARN|AMM|-|Partner is lost (timed out) for interface 1/1/25 LAG sport: 3. State: Expired State - Defaulted State
2019-02-13:09:50:27.351755|lacpd|1321|LOG_INFO|AMM|-|LAG 100 State change for interface 1/1/25: Actor state: ALFOE, Partner state PLFOEX
2019-02-13:09:50:27.351699|lacpd|1309|LOG_INFO|AMM|-|Partner is detected for interface 1/1/25 LAG 100 : 65534,00:00:00:00:00:00. Actor state: ALFOE, partner state PLFOEX
2019-02-13:09:50:27.349862|lacpd|1321|LOG_INFO|AMM|-|LAG 100 State change for interface 1/1/25: Actor state: ALFOX, Partner state ASIOEX
2019-02-13:09:50:24.351619|lacpd|1321|LOG_INFO|AMM|-|LAG 100 State change for interface 1/1/25: Actor state: ALFNX, Partner state ASIOEX
2019-02-13:09:50:24.349745|lacpd|1321|LOG_INFO|AMM|-|LAG 100 State change for interface 1/1/25: Actor state: ALFN, Partner state ASIOEX
2019-02-13:09:48:56.349706|lacpd|1321|LOG_INFO|AMM|-|LAG 100 State change for interface 1/1/25: Actor state: ALFN, Partner state ALIOEX
2019-02-13:09:48:54.386483|lacpd|1321|LOG_INFO|AMM|-|LAG 100 State change for interface 1/1/25: Actor state: ALFO, Partner state ALIOEX
2019-02-13:09:48:54.386406|lacpd|1309|LOG_INFO|AMM|-|Partner is detected for interface 1/1/25 LAG 100 : 32768,44:31:92:2b:30:00. Actor state: ALFO, partner state ALIOEX

 

I logged a ticket to HPE and not much progress there, they suggested to disable stp on the HP5500 side but it's the same. Negotiation start and then port is blocked.

 

Any idea why this might happening?

I'm not familiar with the 5500 but have you tried the lacp mode to active as well instead of dynamic? Algo, looks like you are trunking on the 5500 lag but on the 8320 is an access port, you need to set bith side either trunk or access

+1 about vlan mode...

I have the same issue with 8325 and 2900. Did you find resolution to the issue?

 

Something that i found is that 10Gig gbic link works just fine. Issue is when i try to use 1Gig gbic. It seems that port at 8325 is up but at 2900 it does not come up because 8325 blocks it.

I am seeing the same issue with LAG between an Aruba 8325; 10.03.0010, and several different CISCO switches, in a very basic LAG setup.

Connecting the same CISCO switches to ArubaOS devices via LAG works fine, so it seems that the Aruba 8325 is the cause of the problem.

 

Just got this resolved:

 

First you need to force 2900 LACP ports to 1000-full dublex. Port won't come up without this. Also don't use bottow row for 1gig MM gbic's they won't work proberly (least for me they didn't). Anyway after forcing 2900 ports to 1000-full dublex and only using ports 1/1/1-1/1/2, 1/1/4-1/1/5 etc everything works fine.

For 8325, when you use 1G OPTICAL xcvrs, you also have to force the port to duplex-full. The auto-negotiation is non-existent when using 1G OPTICAL. (it's not necessary for 1Gig-T J8177D).

Thanks, I'll give that a try :-)

 

Update:

I set the interfaces on the 8325 to 1000-full  (no change to the CISCO switch as there ís no speed option on the 1G SFP ports).

LAG doesn't come up, no change in behaviour.

 

8325 keeps logging entries similar to this one:

2019-08-22T10:45:51.240953+00:00 testing-8325 lacpd[1574]: Event|1321|LOG_INFO|AMM|-|LAG 1 State change for interface 1/1/13: Actor state: ALFOE, Partner state PLFOEX

What puzzles me is that the 8325 apparently sees the CISCO link partner as "lacp passive" (partner state P....)
CISCO switch is definitely in LACP active mode.

I'll open up a support case for this.

Greetings, better if you open a new dedicated thread about your issue...this thread was initially used to report issues between an ArubaOS-CX's VSX LAG and a Comware based Switch's LAG, then it was highjacked to report other issues between ArubaOS-CX's VSX LAG and an ArubaOS-Switch (ProVision) based Switch's LAG...and, in the end, also with Cisco...it's not clear where an issue starts and where it ends (and also...apart from initial few configuration lines, nobody else reported the full picture of its own scenario).

 

One suggestion: if you want, please publish sanitized configurations of ArubaOS-CX and Cisco related to LAGs experiencing issues...I'm quite sure interoperability testing was done between Aruba 8000 and Cisco devices with regard to aggregated links interconnection.

Yes, you're right, sorry.

It kind of went into a different direction than intended.

We have found auto-negotiation does not work with J8177D in 8325 switches. But this is not reflected in Aruba documentation anywhere. 

Right. This is missing. Let me follow-up to include this information.

This limitation will be documented in the next revision of the Transceivers guide.