Stack 6850E - Shutdown Primary Unit

[shaps]

Dear,

I discovered a strange behaviour with our stack of OS6850E (4 unit - AOS 6.4.5.474.R02).

The "show stack topology" correctly return Primary & Secondary Unit.

What would be from your point of vue the normal behaviour if we power down the Primary unit?

For your information we have multiple LACP links & a small loop (4 Switches).

The convergence time took around ~4 minutes which looks really long from my point of vue but maybe I'm wrong?

Thanks in advance for your help

Best regards

[devnull]

It may relate to mac-retention (or better change of mac-addresses when primary fails)

please test with
"mac-retention enable" and see whether it converges faster.. It may also be a good idea to do a controlled "takeover" before just powering it down.

[shaps]

devnull,

Thanks for your prompt reply

It's exactly what I was thinking... yesterday I saw that the MAC of the bridge change after the failover.

We have so a complete STP calculation and also saw that some Linkagg failed and need to be re-establish

I will also start a SR to have a position from ALU support

By the way it would be very welcome to have other experience in such scenario... So you think that 6450 have the same behaviour?

Best regards

[shaps]

Also it seems that this command is not supported on AOS6.4.5 ?

Best regards

[devnull]

I think it is pretty standard for stacks to use the mac-address of the primary.
If the primary changes it will use the new primarys mac-address.

6450 in stack-mode behave the same (btw. cisco switches as well).

I would always enable mac retention per default. Only caveat: When you exchange the primary (and reuse it at another place in the network) you will have duplicate MACs in your network which will create a mess..

So: never reuse a ex-primary switch without setting mac-retention to disable on his former stack and verify that the mac has changed. (and reenable it afterwards)

Linkaggs use the MAC-Address for LACP, so if the MAC changes the LACP will be gone until it renegotiates. Therefore STP recalculations take place.

edit: "mac-rentention status enable" was at least mentioned in 6.6.3 Manuals also in 6.4.5 manuals

[shaps]

Hi devnull,

For your information we did a new redundancy test yesterday evening. Before the test we added the mac-retention command

Convergence is now ~1sec inspite of 3 minutes !

Thanks again for your help !

We did the same test on 2 POD (6900-X40) and result is also perfect (no packet lost)... By default the command is enabled on 6900 when you create a Virtual Chassis

Thanks again !

[devnull]

Thanks for feedback..