Ignoring the outstanding bug with ACP for a second.
I have a lot of customers who are now putting in DS4243 disk shelves, and the question always comes up about the ACP configuration. On the 3100 series, PCI cards and GigE ports are at a premium, so its not always practical to assign a single dedicated port to ACP.
The admin guide states that the ACP can be connected through a switch. Reading between the lines, does this mean that if I have an existing VIF and I create a VLAN for ACP, and then VLAN other traffic off separately? I could then make up for the lack of of ports on the filer by piggy backing ACP traffic (fairly low bandwidth I'd imagine) onto an existing connection (which may also be quite low bandwidth) and then onto switch and make full use of ports on a switch stack.
I can't answer authoritatively here but I believe the answer is "no....ACP really should have its own dedicated port if you do it". Given ACP isn't required though, I'd say that if you can't dedicate a port for it, you can just leave it out of the equation.
If you are configuring ACP for disk shelves attached to an active/active configuration, you must supply the same ACP domain name and network mask for both systems.
Attention: Do not connect the ACP port to a routed network, and do not configure switches or hubs between the ACP port and the designated Ethernet port. Doing so is not supported and will cause interruptions in service.
Once you select a domain name and network mask for the interface, the ACP processor automatically assigns IP addresses for traffic in the ACP subnet. This assignment also includes the storage controller network interface that you selected for ACP traffic.
Do we have conflicting docus here? 'cause you mentioned that somewhere it says a switch can be used.
(BTW - what's the ETA on the bug resolution, anyone know?)
Looks like 7.3.2P7 has been released to address this issue, but unfortunately 7.3.3 (which has now gone GD) has not had this fix wrapped up or a patch release available yet.
I'll have to dig out the docs I was referring to, they may have been updated already. Seems odd that you can quite happily push enterpise storage traffic for high end systems over the network, but not ACP management traffic due to potential disruption!!!