Specification
Visit
http://www.hp.com/rnd/products/switches/ProCurve_Switch_2900_Series/overview.htm for more detailed information on the 2900 Switch familiy. Two interesting manuals are
Multicast and Routing Guide as well as
Advanced Traffic Management Guide
Ports
The Switch is available with 24 oder 48
GigabitPorts on the frontpanel, <3.7µsek latecy.
The Back is equiped with:
- 2 CX4 10-GbE ports
- 2 X2 ports (10Gbit depending on installed tranciever)
A CX4 (802.3ak) link Cable is supplied with each 2900.
The latency is <2.1µsek for 10Gbit.
TODO
Restart
To restore
FactoryDefaulty:
- Press both the Clear an the Reset button simultyniously.
- release the Resetbutton and wait untill the SelfTest LED starts to falsh.
- release the Clearbutton.
Configuration
By default the switch has no IP/Subnetmask set. Our Sample has Hardwareaddress 00:16:b9:0c:ac:40 recives 192.168.1.13 via Dhcp. The web interface is forwarded to 130.75.117.77:11080.
Configuration can also be made using
- telnet
- SNMP
- A SNMP based Management Software (ProCurve Network Management)
- Web interface
Try an snmpwalk to get all possible Values.
snmpwalk -c public -v1 192.168.1.13 .1.3.6.1.4.1.11
enterprises.11 (enterprises = .1.3.6.1.4.1 in the MIB tree) is the vendor HP.
Every Tree-location on this page is beond this root
x stands for 1-52, the Numbers of the ports.
Location |
Name |
Value |
Comment |
2.14.11.5.1.1.6.0 |
Base MAC Address(?) |
Hex |
2.14.11.5.1.7.1.3.1.1.4.x |
hpSwitchPortAdminStatus |
|
(enabled) 1 = yes, 2 = no |
2.14.11.5.1.7.1.3.1.1.3.x |
hpSwitchPortDescr |
String |
DisplayString "1-48" and "A1-A4", but seems to be Read-only |
2.14.11.5.1.7.1.3.1.1.2.x |
hpSwitchPortType |
|
119:1Gbit, 2:X2, (122:none?) |
2.14.11.5.1.7.1.3.1.1.8.x |
hpSwitchPortTrunkGroup |
INTEGER |
Trunk Group (1-36), = disabled |
2.14.11.5.1.7.1.3.1.1.13.x |
hpSwitchPortTrunkType |
INTEGER |
5 none, 4 LACP, 1 trunk |
2.14.11.5.1.7.1.3.1.1.14.x |
hpSwitchPortTrunkLACPStatus |
INTEGER |
1 = disabled, 2 = active, 3 = passive (Defines LACP status if TrunkType is LACP) |
2.14.11.5.1.7.1.3.1.1.15.x |
Port status |
|
4 = down, 3,2 = up, 1(10GbitPorts) |
2.14.11.5.1.7.1.15.3.1.3.x |
Type |
|
119:1Gbit, 2:10Gbit, 54:Trunk |
2.14.10.2.14.1.2.1.2.1.2.x.y |
some port setup |
|
x:port,y:1-4 |
2.14.11.5.1.7.1.14.2.1.x |
MacAddrees of port |
|
|
Trunked Devices appear as 288+#, 289 beeing trunk1, in the Mib Tree.
You can walk this MIBs very easy via
telnet:
walkMiB NAME
telnet
Usage
- Type help to get available commands. For further information type command and then '?'
- Type config to move to the 'global configuration level' with more commands
- One can also use the menu-style configuration. Type menu
- Using get/set/walkMIB one can modify MIB Entrys. setMib NAME -type value, Type can be i,s...
# Example: Howto trunk ports via CLI
# Trunk ports 25 to 28 via lacp to group named trk10
trunk 25-28 trk10 lacp
# Delete trunk
no trunk 25-28
# Build a dynamic trunk group:
interface 25-28 lacp active
#configure
#vlan 1 tagged 4,5,6
Allow ports 4,5,6 to recive tagged packets from
VLAN 1
Jumbo Frames have to be enabled for every
VLAN.
#configure
#vlan 1 jumbo
Enables Jumbo Frames for the Default
VLAN (1) and Check via
#show vlans
If no Networkconnection can be established, one can use the serial console to manage the Switch.
Run
minicom -s Set
BaudRate to 9600 8N1, disable "Hardware Flow Contorl" and Inizialize the connection
STRG-A M.
If the Swich has accidentialy set to a wrong boot image. Type
boot , wait and choose "Primary firmware Image".
ProCurve Network Management
A 30-day trial Version was included. There is a Windows install.exe..
Web Interface
One can change the ports configuration. Watch out for the button
Modify selected Ports
Tests
10Gbit Ports
- Connecting both 10Gbit Cables creates a huge amount of network traffic. The whole System is crashing.
-->Enable Trunking for Port A2,A3 and it works.
Stack
If two or more Switches are installed, they automatically share their Stacks.
Go
Telnet->MainMenu->Stacking to enable/disable this feature.
Channelbonding
visit
channelbonding for detailed Information on the linux bonding module.
Ports connected to the same box must be trunked together.
round-robin should work out-of-the-box, wheras
LACP requires the trunk-type (hpSwitchPortTrunkType) to be set to LACP.
Internal Algorithms
The Trunking loadbalancing on the recive side chooses the port based on a hash on Src/Dest-MAC-addresses.
As a result Traffic from one Source MAC-Address to one Destination MAC-Address can only go throug one port of a Trunk! In the Linux Channelbonding Module all interfaces in a bond chare the same MAC-Address. As a result the maximum throughput between two node over a Switch is limited to the speed of
one single interface at the recive node.
A solution would be to use
VLANs in a particular way.