PowerConnect 62xx User's Manual Stacking is WRONG?

Question

We are evaluating the 62xx series stacking capabilities, and we came over something inside the User's Manual that we cannot understand as being correct. Here's what the manual says: 'In Figure 2-3, the stack has the following physical connections between the switches: • Unit 1 and Unit 2 are connected through the XG1 ports on each switch. • Unit 2 and Unit 3 are connected through the XG2 ports on each switch. • Unit 1 and Unit 4 are connected through the XG2 port on Unit 1 and the XG1 port on Unit 4. The resulting ring topology allows the entire stack to function as a single switch with resilient fail-over capabilities.' How come this is resilient or a ring? If we loose Unit 1, we will end with Unit 4 also lost. If we loose Unit 2, we will also loose Unit 3, etc. Or the cascade modules would still work even on dead switches? We suppose the ideal stacking solution would be: Unit 1 -> Unit 2 -> Unit 3 -> Unit 4 -> Unit 1 again We think we could do something like stated above, no? Can someone explain how the example from the user's manual is resilient or is it really wrong? Thank you for your time on this one! Regards, Message Edited by malheiros on 11-06-2007 03:40 PM

malheiros · Answer

Exactly that's the only thing we imagined to be possible (cascade modules work even with damaged/powered off units) on the setup presented by the user manual. Are you sure the cascade modules/cables work even with damaged/powered off units or you just think so its possible?

We have not bought the switches yet, we are going through it's specs and all so we can't run tests like these (yet) :)

Still, remains the question if we can cascade the switches on a true ring configuration/loop like described on our first message.

Thanks mandzo

Best regards,

mandzo · Answer

If you are evaluating 62XXs, did you tried to power OFF UNIT 1? Stacking modules and cables are looking like passive elements, aka backplane. So, the UNIT 4 should still have a connectivity.

mandzo · Answer

We have a stack of 5 switches with full ring of cables as it was described in your post, but this is a production stack and I can't test it. Probably the best way would be to get a conformation of this from the DELL moderator or I could check with a colleague who might have done it in the past.

Dell-Steve A · Answer

Malheiros,

The manual page isn't showing a redundant stack-so in the event of switch-2 going down, switch 3 will be isolated. I'm looking at fig 2-7, and the diagram is not showing a redundant ring. (It's possible that the manual has changed since you posted this, as the current one now includes the 6224F-the URL I'm using is http://support.dell.com/support/edocs/network/PC62xx/en/UG/PDF/ug_en.pdf)

The proper setup for redundancy is

XG1 on switch 1 -> XG2 on switch 2

XG2 on switch 2 -> XG2 on switch 3

XG1 on switch 3 -> XG1 on switch 4

XG2 on switch 4 -> XG2 on switch 1

For resiliency, a backup master is selected and the configuration is copied over to that switch in the event that the master fails.

Message Edited by Dell-Steve A on 11-07-2007 08:21 AM

malheiros · Answer

Thats exactly what we were thinking. That manual is leading to some confusion since it states that's a resilient configuration, which is not. And the picture shows two empty slots on units 3 and 4, that doesn't make any sense. Someone should review that and correct that issue, since many ppl will get confused with that as well.

As you can see me and mandzo started to think that the switches would forward traffic on the stacking modules even if they are damaged or powered off, like a backplane. No one that buy those kinds of switches would/should stack anything without redundancy, so the manuals should have only redundant stacking examples.

Just a little correction on your suggestion (just a typo), switch 2 should use XG1 for the first stacking connection with switch 1, so it should be like this:

XG1 on switch 1 -> *XG1* on switch 2
XG2 on switch 2 -> XG2 on switch 3
XG1 on switch 3 -> XG1 on switch 4
XG2 on switch 4 -> XG2 on switch 1

So to make something like this, you need 4 stacking modules with 4 cables and you are good to go.

I want to thank you guys for helping us solving that issue, we really appreciate it. :)

Best regards,

mandzo · Answer

Steve, Do you have a link for the updated version of CLI Reference Guide? In the latest revision of the User Guide, they started talking for a 'STANDBY' unit, which wasn't mentioned before. Also, the latest firmware ver. 2.0.0.12 doesn't have any option to specify a STANDBY unit. The stack has an option to specify the master: Admin Management Preference, which I am using. There is another setting a 'Hardware Management Preference', which states 'Unassigned'. Is this assignment done by predefining the master unit and not using the election process? Thanks

4567893 · Answer

As an aside, I found out that if the master unit in the stack dies, the entire stack reboots to elect a new master.  That may or may not be a reason not to purchase these for its stacking capabilities in a production environment.

mandzo · Answer

Hi 4567893 , which firmware version were you running when the master died? And maybe this is the reason why DELL started to talk for a "STANDBY" unit to realy achieve resilent operations?

Message Edited by mandzo on 11-09-2007 08:55 AM

4567893 · Answer

At the time I tested, the switches were on 1.0.4.3.  That was while I still had them in the lab and not in production.  Now that they are in production, I've since upgraded to 2.0.0.12, but I didn't see anything in the release notes that stated they changed that behaviour.  I could have missed it.

Networking General

Was this post helpful?