Link together a series of 2510G-48 for one subnet

[Steve Ireland (the othe]

We have 7 x HP ProCurve Switch 2510-48G (J9280A), 14 x HP Procurve miniGBIC/1xGENet SX LC (J4858C), 7 x 2 m LC-LC Multi-Mode OM3 Fibre Channel Cable (AJ835A), 1 x HP ProCurve 2520G-24-PoE Switch (J9299A) but very little experience.

We have rack mounted all switches, two at remote locations.

Daisy-chaining them, each switch is connected to its two partners by GBICs in ports 47 & 48 with the two exceptions of a 2510G-48 (SwitchFirst - that is also connected to a remote switch over an extra GBIC/fibre) and the 2520G (SwitchPOE - the last switch in the line).

For the most part there is only one link between each switch yet that link is enabled as a Trunk.

To illustrate:
SwitchFirst/46/Trk7 <-> SwitchAnnex/48/Trk7
SwitchFirst/48/Trk1 <-> SwitchSecond/47/Trk1
SwitchSecond/48/Trk2 <-> SwitchThird/47/Trk2
SwitchThird/48/Trk3 <-> SwitchFourth/47/Trk3
SwitchFourth/48/Trk4 <-> SwitchFifth/47/Trk4
SwitchFifth/48/Trk5 <-> SwitchSixth/47/Trk5
SwitchSixth/48/Trk6 <-> SwitchPOE/47/Trk6

All we really wanted was to run one 10.2.0.0/16 Subnet (for the moment) over all switches and - in so far as is possible - run them all as one great, big switch. For Ghosting, we might temporarily VLAN segments of the switches (tagging the trunks) when we want to "Directed Broadcast" Ghost sessions (as Multicasts are killed by the switches - we'd actually love to MultiCast). Finally, we need to Aggregate Ports (I am guessing with more static Trunks or LACP) to four NICs on a Dell PER710 so that students doing video editing can pull Gigabytes of data from one server with a SAN behind it full of high speed disks (to keep it simple, we figured we'd put all the video editing PCs on the same switch as the four NICs for the server.

Spanning Tree and IGMP are not enabled (I think).

We have stacked the switches under one IP but, every now and again, connectivity with that IP becomes untenable. Pings time out or go up to 650ms or similar. We are therefore sometimes (sometimes it's fine!) unable to telnet/HTTP to switches. Right now, we can't reach the switches/stack so I can't do "sh tech".

Are we doing this correctly? 'Cause it doesn't feel like we are. We think the Trunks are the issue (are we naming/defining them incorrectly?).

If we obliterate the current configuration, can you suggest how we should have this configured? (Also, as an aside... should I do away with the single GBIC connections between switches and use the copper interfaces paired/tripled as trunks instead? Were these GBICS a waste of money...?)

[Steve Ireland (the othe]

I removed all trunk designations on all switches. They are now simply linked by the GBIC fibre connections. So far so good. Communications seems fine.

I guess I'll carry on figuring this one out on my own...

[Pieter 't Hart]

Initially you had trunks with only a single fiber port?

>>> Were these GBICS a waste of money <<<
Well, fiber is supposed to pick up less interference then copper.
As for the distance, copper should have worked too.
Only the long-distance switches really benefit from the fiber connection.

Maybe you can connect these remote switches with an LACP trunk of two fibre links, and connect the switches that are close to eacother with two copper links?

>>> We have stacked the switches under one IP<<<
that means one of the switches actually coordinates the management, you may pass multiple links before the switch that serves this management addrress is reached.

you can create a separate management-vlan to separate management traffic from other data.