Dell EMC SmartFabric OS10 User Guide Release 10.5.0

Example: Manual rebalance trigger

This section describes how to use the manual rebalance CLI command to rebalance the load when the system is in an unbalanced state.

OS10#re-balance npg sessions vfabric 100

Fabric Id 100 state before rebalancing

Uplink   FLOGI        FDISC          Load              Speed         Excess
Intf                                                           (Gbps)         Load
-------------------------------------------------------------------------------------
Fc 1/1/1          1            9              10                8             7
Fc 1/1/2          3            3              6                 16            0
-------------------------------------------------------------------------------------
                  4            12             16                24            7
-------------------------------------------------------------------------------------

Session Re-distributions:

16 Session Re-distribution(s)
--------------------------------------------------------------------------------------------
Node WWPN                       From Uplink Intf        To Uplink Intf      No.of sessions
--------------------------------------------------------------------------------------------
20:01:d4:ae:52:1a:ee:54         Fc 1/1/1                    Fc 1/1/2        10
21:01:d4:ae:52:1a:ee:54         Fc 1/1/2                    Fc 1/1/1        2
22:01:d4:ae:52:1a:ee:54         Fc 1/1/2                    Fc 1/1/1        2
23:01:d4:ae:52:1a:ee:54         Fc 1/1/2                    Fc 1/1/1        2

Fabric Id 100 expected state after rebalancing

Uplink     FLOGI          FDISC        Load           Speed            Excess
Intf                                                           (Gbps)           Load
--------------------------------------------------------------------------------------
Fc 1/1/1            3              3            6              8                1
Fc 1/1/2            1              9            10             16               0
--------------------------------------------------------------------------------------
                    4              12           16             24               1 
--------------------------------------------------------------------------------------

You can use manual rebalancing when you:

Add new FC uplink to a balanced system

Consider a topology with the following structure:

• NPG switch with two FC uplinks (fc 1/1/1 and fc 1/1/2) of the same speed (16G)
• Ports connecting to both FCoE and FC end points (eth 1/1/54, eth 1/1/55, fc 1/1/9 and fc 1/1/10)

All the end points (servers) are logged in to the storage through the NPG switch. One FLOGI session is associated with each server.

The NPG switch has four sessions that need to be mapped between two FC uplink (fc 1/1/1 and fc 1/1/2).

The NPG switch maps two sessions from eth 1/1/54 and eth 1/1/55 to fc 1/1/1, and two sessions from fc 1/1/9 and 1/1/10 to fc 1/1/2 to maintain a balanced state.

When you add two more FC uplinks, fc 1/1/3 and fc 1/1/4 in the NPG switch, the system becomes unbalanced. Issue a manual rebalance CLI command to move the system back to a balanced state, with each FC uplink (fc 1/1/1, fc 1/1/2. fc 1/1/3 and fc 1/1/4) having one session each.

Receive session log out (FLOGO or CVL) from the end points

Consider the NPG switch with:

• two FC uplinks (fc 1/1/1 and fc 1/1/2) of the same speed(16 G)
• four ports (eth 1/1/54, eth 1/1/55, fc 1/1/9 and fc 1/1/10) connecting the FC and FCoE end points

Each end point has one session associated with it. The NPG switch maps two sessions to each FC uplink to balance the system.

Consider both the FC end points sending a session termination request (FLOGO) to the NPG switch. The system closes the sessions in fc 1/1/10, and the system moves to an unbalanced state.

When you bring two more FC uplinks fc 1/1/3 and fc 1/1/4 in the NPG switch, the system becomes unbalanced. Issue a manual rebalance CLI command to move the system back to a balanced state, with each FC uplink (fc 1/1/1, fc 1/1/2. fc 1/1/3 and fc 1/1/4) having one session each.

Now consider both the FC end points sending a session termination request (FLOGO) to the NPG switch. The system closes the sessions in fc 1/1/10, and moves to an unbalanced state.

When you issue the manual rebalance command, the system redistributes the session so that the FC uplinks (fc 1/1/1 and fc 1/1/2) have one session each.

Receive Fabric Discovery Request (FDISC) from an end point

Consider the NPG switch with:

• two FC uplinks (fc 1/1/1 and fc 1/1/2) of different speed (8 G and 16 G)
• two ports (eth 1/1/54, eth 1/1/55) connecting the FCoE end points

Each end point has one session that is associated with it. The NPG switch maps one session to each FC uplink to balance the system.

Consider the end point connected to eth 1/1/55 establishes four more Fabric Discovery Sessions (FDISC). The FDISC sessions are mapped to the same FC uplink, which has the FLOGI of the end point (fc 1/1/1) and the system moves to an unbalanced state.

When you issue the manual rebalance command, the system redistributes the sessions and the uplink 1/1/1 has one session from one of the end points. The remaining five sessions from the other end points are mapped to fc 1/1/2.

This makes the system better where in the link with the maximum speed holds more sessions than the link with minimum speed.