2

I have the following items:

  • Primergy BX600 MMB3
  • 10 x BX620 S4
  • 2 x Switch Blade (PG-SW107)
  • QNAP (TS-EC1279U-RP – QNAP) (4 x Gigabit Connections)
  • HP Pro curve Switch (2510g-48 (JA9280-A))

Patch Panel connecting all ports from the (2 x Switch Blade (PG-SW107)) to the patch panel

What I would like to do:

I have setup everything except the network switches. Some advice on how and what is the best way. It has been some time since I have setup a switch.

Here are the switches: 2 x Switch Blade (PG-SW107)

  • Total ports - 44 (Physical, internal and intern connection)
  • Down link - 30 (Internal server blade connection)
  • Up link - 12 (12 uplink ports for external connection)
  • Interlink - 2 (Inter-switch-blade connection)

The switch has 44 ports in total; 30 down link ports for internal server blade connection, 12 uplink ports for external connection, and two interlink ports for inter-switch-blade connection. It is compatible with Fujitsu's SRS secure switches in operability. Up to two switches can be installed in the network blade slots in the chassis rear.

Here is an example of 2 x Switch Blade (PG-SW107)

enter image description here

Here is the menu of the switch:

enter image description here

Issues

At present when I plug in BOTH PG-SW107 it stops all the network communication.

My questions:

  1. Is it recommended to create a VLAN for just the iSCSI ( 2 of the 4 Gbit connections). Use the remaining 2 to connect to the HP switch for file access?
  2. From the above menu, how would I connect port 43 and 44 as the direct link?
  3. How would I get both switched to work as in the picture near the top? I have looked online and I could not really find and information about this, but maybe I am looking in the wrong place.
  4. Would I connect 1 gbit to to one switch and the other GBIT to the other switch all on the same VLAN for ISCSI and I guess a level of redundancy?

Thank you for your help!

Arthor
  • 195
  • 4
  • 18
  • I am not sure why I have been down voted for a perfect good question and also a valid one. – Arthor Jul 28 '12 at 00:36
  • My guess would be because it seems like your question could be answered by reading the manual. – HopelessN00b Jul 28 '12 at 14:16
  • @HopelessN00b I have been through the manual and unfortunately it is not that details. Even more so the Fujitsu forum is really not that active. I guess my first step is how to bond the internal NIC to each blade. – Arthor Jul 28 '12 at 15:38
  • I have made a chat room if any one is willing to help a little: http://chat.stackexchange.com/rooms/4289/setup-blade-server-switch – Arthor Jul 28 '12 at 22:09
  • @Arthor don't do that, never create chat rooms just because you fancy it, it's pointless. And haven\t you asked a very similar question to this a week or two back that I answered? – Chopper3 Jul 29 '12 at 12:56

2 Answers2

2

1). Yes, it makes sense to separate your iSCSI traffic onto a separate vlan from your data. Dedicate one or two blade uplinks per server to that. I'd probably use two uplinks, leveraging one port from each switch blade for redundancy and iSCSI MPIO (e.g. port C and F from each server would be dedicated iSCSI).

2). Ports 43 and 44 look to be an internal stacking link/cross-connect between the b;ade enclosure's switch modules. This is standard. You don't need to do anything to activate it.

3). That may be a manual or blade management software question. To clarify, this server is a Fujitsu Primergy BX600 MMB3. I've never heard of it before, but there have to be other installations out there.

4). I'm not sure I understand your question here.

ewwhite
  • 197,159
  • 92
  • 443
  • 809
  • Hello @ewwhite, I will be logged in the chat all day. I understand what you are saying about 1. Thank you. 2. I agree what you are saying however when I but both switch blade into the chassis and they come online. They kill the network. I see all the lights going crazy and then I cannot even access the management board. Could it be creating an infinity loop or something. 3. Yes, that is correct. 4. when both blade switch's are working I would connect from the qnap one gbit cable to the blade sw - 1 and the other gbit to blade sw - 2. to have redundancy – Arthor Jul 29 '12 at 13:31
  • For #2. Make sure you have "Spanning Tree" enabled on your ProCurve switch. – ewwhite Jul 29 '12 at 13:39
  • This happens without and connection to the pro curve.. I assume both blade switches and ports must have spanning tree enabled. – Arthor Jul 29 '12 at 13:42
  • I would defer to the manufacturer or their support. I'm not familiar enough with the Fujitsu server solutions to be able to provide better advice. – ewwhite Jul 29 '12 at 13:43
  • I totally agree with you however the forums are not active at all and the manuals does not give example setups... I cannot find anything about setting up the blade switch's hence here. Okay. I will try. – Arthor Jul 29 '12 at 13:45
  • I enabled port 43 and 44 (inter link) with the spanning tree protocol on both switches but still the same. So I am sure I am missing something. – Arthor Jul 29 '12 at 14:21
0

For those in a similar situation, try the flowing as it works well.

  1. Disable BOTH inter-link power between both blade switches
  2. Disables STP as this is not needed
  3. Create a trunk with 6 ports on blade switch 1 to the Pro Curve
  4. Configure the pro cure with the appropriate trunk (Do not put all the connection in one module location on the pro cure, spread them out to distribute the load.
  5. Do the same for the same above on blade switch 2.

Here you will be able to setup vMware with a good level of resilience. Unfortunately the blade switches are NOT be stacked.

Your SAN/NAS.

  1. Connect one connection to Blade switch one (192.168.10.1)
  2. Connect one connection to blade switch two (192.168.11.1)
  3. Use Round Robin + iSCSI and you will have 2 path via iSCSI and with the Blade switches.

Thanks all for your support.

Arthor
  • 195
  • 4
  • 18
  • This worked much better – Arthor May 01 '13 at 13:53
  • Indeed, I believe you should not interconnect two switches that are supposed to provide redundancy in the chassis. We for example have two completely independent fibrechannel fabrics instead of a full mesh or the like. As far as the network is concerned, I believe there is no need to interconnect the blade switches since they represent the edge and the network is probably redundant in its core. Interconnecting the blade switches would only generate a more complicated STP computation and no additional benefit IMHO. – Marki Dec 20 '14 at 19:22