Reset loop-protect question for Virtual Connect

[chuckk281]

Question on Network Loop Protection from Peter:

*****************

Has anybody  done this yet and can tell me if there is any disruption to the network on any other server in the enclosure ?

Page 88 and 89 from the manual says you need to do it when a bay has a network loop detected … but doesn’t tell me if it needs downtime for the other servers in the enclosure or not ?

http://bizsupport1.austin.hp.com/bc/docs/support/SupportManual/c03128009/c03128009.pdf

Network loop protection

To avoid network loops, Virtual Connect first verifies that only one active uplink exists per network from the

Virtual Connect domain to the external Ethernet switching environment. Second, Virtual Connect makes sure

that no network loops are created by the stacking links between Virtual Connect modules.

• One active link—A VC uplink set can include multiple uplink ports. To prevent a loop with broadcast

traffic coming in one uplink and going out another, only one uplink or uplink LAG is active at a time. The

uplink or LAG with the greatest bandwidth should be selected as the active uplink. If the active uplink

loses the link, then the next best uplink is made active.

• No loops through stacking links—If multiple VC-Enet modules are used, they are interconnected using

stacking links, which might appear as an opportunity for loops within the VC environment. For each

individual network in the Virtual Connect environment, VC blocks certain stacking links to ensure that

each network has a loop-free topology.

Enhanced network loop protection detects loops on downlink ports, which can be a Flex-10 logical port or

physical port. The feature applies to Flex-10 logical function if the Flex-10 port is operating under the control

of DCC protocol. If DCC is not available, the feature applies to a physical downlink port.

Enhanced network loop protection uses two methods to detect loops:

• It periodically injects a special probe frame into the VC domain and monitors downlink ports for the

looped back probe frame. If this special probe frame is detected on downlink ports, the port is

considered to cause the loop condition.

• It monitors and intercepts common loop detection frames used in other switches. In network

environments where the upstream switches send loop detection frames, the VC Enet modules must

ensure that any downlink loops do not cause these frames to be sent back to the uplink ports. Even

though VC probe frames ensure loops are detected, there is a small time window depending on the

probe frame transmission interval in which the loop detection frames from the external switch might loop

through down link ports and reach uplink ports. By intercepting the external loop detection frames on

downlinks, the possibility of triggering loop protection on the upstream switch is eliminated. When

network loop protection is enabled, VC-Enet modules intercept the following types of loop detection

frames:

o PVST+ BPDUs

o Procurve Loop Protect frames

When the network loop protection feature is enabled, any probe frame or other supported loop detection

frame received on a downlink port is considered to be causing the network loop, and the port is disabled

immediately until an administrative action is taken. The administrative action involves resolving the loop

condition and clearing the loop protection error condition. The "loop detected" status on a port can be

cleared by one of the following administrative actions:

• Restart loop detection by issuing "reset" loop protection from the CLI or GUI

• Unassign all networks from the port in "loop detected" state

The SNMP agent supports trap generation when a loop condition is detected or cleared.

Virtual Connect provides the ability to enable or disable network loop protection. The feature is enabled by

default and applies to all VC-Enet modules in the domain. Network loops are detected and server ports can

be disabled even prior to any enclosure being imported.

A loop-protect reset command resets and restarts loop detection for all server ports in a “loop-detected” error

condition.

***************

Reply from Chuck:

**************

It will/should only shut down the downlink port connected to the server causing the loop.  It is essentially a protection against customers who have bridged their internal server nics and have caused a loop. No other ports will be affected.

Any server networks using that downlink port (uplink port from the servers’ perspective) will be inaccessible until the bridge is removed and the reset loop command is issued.

****************

Comments?