LACP Timeout with Virtual Connect and A-Series 5120 IRF connected switches

[Manfred M.]

We have 2x c7000 enclosures in different IT Rooms stacked together and LACP Uplinks to two pairs of 5120 Switches connected together in an IRF-chain at both sides.

In addition to the VC-Stacking the sites are connected with 2x 10G LACP between the switches.

There are 2x 10G Uplinks between the VC Module 1 on one site and VC Module 2 at the other site and the 5120 Switches on both sites. All Uplinks are active in Virtual Connect Manager.

The switches are permanently showing LACP Timeouts (Flag B - see below) - no matter if we configure 'lacp short period' on the 5120 interfaces or use the default setting of the long period of 30s. Even if we shutdown one of the uplinks in a Bridge Aggregation Group the problem still exists.

[core]display link-aggregation verb Bridge-Aggregation 3

Loadsharing Type: Shar -- Loadsharing, NonS -- Non-Loadsharing
Port Status: S -- Selected, U -- Unselected
Flags:  A -- LACP_Activity, B -- LACP_Timeout, C -- Aggregation,
        D -- Synchronization, E -- Collecting, F -- Distributing,
        G -- Defaulted, H -- Expired

Aggregation Interface: Bridge-Aggregation3
Aggregation Mode: Dynamic
Loadsharing Type: Shar
System ID: 0x8000, 3822-d60a-e837
Local:
  Port             Status  Priority Oper-Key  Flag
-----------------------------------------------------------------------------
  XGE1/2/1         S       32768    5         {ACDEF}
  XGE2/2/1         S       32768    5         {ACDEF}
Remote:
  Actor            Partner Priority Oper-Key  SystemID               Flag
-----------------------------------------------------------------------------
  XGE1/2/1         17      1        4         0x1   , 101f-7478-a37f {ABCDEF}
  XGE2/2/1         18      1        4         0x1   , 101f-7478-a37f {ABCDEF}

We do have Performance Problems in that configuration - so we have to find out what's going wrong here.

Any ideas?

Hint's are very welcome!

Manfred

[Manfred M.]

Additional remark:

Even on a Screenshot in the 'Virtual Connect and HP A-Series switches IRF Integration
Guide' on page 18 I have found this 'B' flag...

I have found ot that we have a very similar 'active-active' configuration as described on scenario 2 on page 3 of that integration guide.

The only difference is, that we have 2x IRF Rings which are connected together with a LACP Bridge Aggregation Group.

[Manfred M.]

In the meantime I have found a Support Information (emr_na-c02065408):

HP Virtual Connect Ethernet Module for c-Class BladeSystem - LACP Timeout Value Information

Here is a short explanation on how the LACP timeout value works on the Virtual Connect module.

Details

IEEE standard:

43.4.13

Periodic Transmission machine

The Periodic Transmission machine establishes the desire of the Actor and Partner to exchange periodic LACPDUs on the link in order to maintain an aggregation, and establishes how often those periodic transmissions should occur. Periodic transmissions will take place if either participant so wishes. Transmissions occur at a rate determined by the Partner; this rate is linked to the speed at which the Partner will time out received information.

The Virtual Connect module will always try to negotiate a short timeout value. This means it will request LACPDUs every second. If the Virtual Connect does not receive a LACPDU after 3 seconds, it will timeout and suspend the LAG.

We have configured the short Period at the A-Series Switches (which is the default setting) - for me it's still unclear, if the Timeout failure is a problem or not - we did not see any suspended LAG...