Brocade Communications Systems 53-1002163-02 manual High availability, Hot plug

High availability 5

On an FCOE10-24 this mapping is as follows, assuming that the FCOE10-24 blade is in slot 4:

•FCoE port 4/0 maps to TE port 4/0

•FCoE port 4/1 maps to TE port 4/1

•FCoE port 4/23 maps to TE port 4/23

High availability

Fabric OS v7.0.0 ensures high availability of FCoE storage traffic going through FCOE10-24 blades on the DCX and DCX-4S platforms. If the Control Processors (CPs) on the DCX or DCX-4S are in a redundant state and the active CP fails, the standby CP takes over the control of the system. Failover from the active CP to the standby CP is non-disruptive to both FC and FCoE traffic.

Hot plug

The procedure for hot-plugging an FCOE10-24 is similar to any other blade supported on a DCX or DCX-4S. When diagnostics are running on a blade, a slotpoweroff operation is not allowed. This behavior is similar to what has been supported in the previous FOS releases.

When a core blade on a DCX or DCX-4S is removed, the traffic flowing through the ASICs in that blade is re-routed to use the ASICs in the other core blade in the chassis. This increases the congestion, but the switch continues to operate in an over-subscribed mode. When both the core blades are removed, then all the blades in the chassis are disabled.

Control Processor failover

When the active CP in a DCX or DCX-4S fails over, the FCoE storage traffic going through the FCOE10- 24 blades are not disrupted. Critical protocol components are brought online quickly to ensure that PDUs are transmitted to the adjacent switches so that the associated protocol sessions are not declared dead.

ATTENTION

Upgrading your firmware from Fabric OS v6.3.x, v6.4.0, or v6.4.1 does not cause FC traffic disruption but does cause FCoE traffic disruption.

Brocade recommends that the protocol timeout values in Table 9 be configured both on the local Brocade switches, as well as neighboring switches, to ensure non-disruptive HA failover.

TABLE 9	Protocol timeout values for non-disruptive failover
Protocol type	Time interval
LACP	Long timeout = 90 seconds (default value)
FIP	Hello Interval = 8 seconds (default value)
RSTP	3 Second Hello (9 second session timeout)
MSTP	3 Second Hello (9 second session timeout)

Full recovery of all services can take up to a minute. No new configuration is allowed for that period. Configuration manager generates a RASLOG message to indicate the availability of the standby CP after recovery to accept new configuration CLIs.

Converged Enhanced Ethernet Administrator’s Guide	33
53-1002163-02