Configuring Shared Ethernet Adapter Failover :

This document describes some general concepts related to Shared Ethernet Adapter

(SEA), the procedures to configure SEA Failover, as well as migrating from an existing

configuration to SEA Failover.

Shared Ethernet Adapter:

A Shared Ethernet Adapter can be used to connect a physical network to a virtual

Ethernet network. It provides the ability for several client partitions to share one physical

adapter. SEA can only be configured on the Virtual I/O server (VIOS) and requires the

POWER Hypervisor and Advanced POWER virtualization feature. The SEA, hosted on

the Virtual I/O server, acts as a Layer-2 bridge between the internal and external network.

Restrictions with configuring SEA Failover:

· It can only be hosted on the VIOS and not on the client partition.

· The VIOS running Integrated Virtualization Manager (IVM) cannot implement SEA

Failover because only one single VIOS can be configured on the P5 with IVM.

· SEA Failover was introduced with FixPack 7, i.e., in Virtual I/O server version 1.2.

so both the Virtual I/O servers need to be at this minimum level.

Requirements for configuring SEA Failover:

· Two Virtual I/O servers, both atleast at FixPack 7, with bridging functionality of

Shared Ethernet Adapters configured on them.

· One SEA on one VIOS acts as the primary (active) adapter and the second SEA on

the second VIOS acts as a backup (standby) adapter.

· Each SEA will have at least one virtual Ethernet adapter with the “Access external

network” flag (previously known as “trunk” flag) checked.

· This adapter on both the SEAs has the same PVID, but will have a different priority

value.

· A SEA in ha_mode (Failover mode) might have more than one trunk adapters, in

which case all should have the same priority value.

· The priority value defines which of the two SEAs will be the primary and which will

be the backup. The lower the priority value, the higher the priority, e.g. an adapter

with priority 1 will have the highest priority.

· An additional virtual Ethernet adapter , which belongs to a unique VLAN on the

system, is used to create the control channel between the SEAs, and must be specified

in each SEA when configured in ha_mode.

· The purpose of this control channel is to communicate between the two SEA adapters

to determine when a fail over should take place.

Steps to configure SEA Failover:

· Check that the P5 is attached to the HMC.

· Both the VIO servers should be at least at version 1.2. The latest FixPack has

some SEA fixes, so we recommend upgrading to the latest FixPack level.

· Each VIO server should have at least one available physical Ethernet adapter

assigned to it.

· Using the HMC GUI, create a virtual Ethernet adapter on each of the VIO servers,

configure the adapter as the trunk adapter, each with a different trunk priority

value.

· The SEA configured with the trunk adapter with the lowest priority value will be

the active (primary) adapter. Set the priority=1 on the adapter with “Access

external network” flag checked on the VIOS that you have chosen to configure as

the primary SEA.

· Set the priority=2 or higher on the adapter with “Access external network” flag

checked on the second VIO server. This VIOS will have the backup SEA adapter.

· Make sure that the PVID and the VID values of the trunk adapter for VIOS 1 and

VIOS 2 are the same.

· Configure another virtual Ethernet adapter (to be used for the control channel) on

each of the VIO servers. These adapters must have the same PVID. Double check

that this PVID is unique to the system, i.e., that no other virtual Ethernet adapter

on the P5 uses this PVID. The following screen shot shows the configuration of

control channel of VIOS 1. The screen shot for configuration of control channel

of VIOS 2 would look the same (however, the slot number may differ).

VIOS 1

· Then on CLI, first on one VIOS and then on the second, run the following

command:

mkvdev -sea physical_adapter -vadapter virtual_adapter -default virtual_adapter\

-defaultid PVID_of_virtual_adapter –attr ha_mode=auto \

ctl_chan= control_channel_adapter

Migrating From Existing Configuration to SEA Failover:

It is possible to migrate an existing SEA configuration on the servers, with NIB on the

client, to SEA Failover with only a short, planned downtime on client LPARs.

A prerequisite is a working DLPAR environment, as some virtual Ethernet adapters will

be added using DLPAR. An adapter added via DLPAR must also be added to the profile,

in order to maintain the configuration after a reboot.

The NIB on the client LPAR can be removed at a later time and can be scheduled at a

different time for each client LPAR. This implies a short downtime on a client LPAR.

Switch configuration from NIB & SEA to SEA Failover:

· Add a virtual Ethernet adapter to each VIOS which will be used as the control

channel for SEA Failover. After the DLPAR operation, add the adapter to the

profile.

· Activate SEA Failover feature on VIOS 1 by running the chdev command to

switch on control channel and ha_mode:

$ chdev -dev ent3 -attr ha_mode=auto ctl_chan=ent4

ent3 changed

$ lsdev -dev ent3 –attr

…

ctl_chan ent4 Control Channel adapter for SEA Failover True

ha_mode auto High Availability Mode True

…

· Add a virtual Ethernet adapter to VIOS 2 which has the same PVID as virtual

Ethernet on VIOS 1, check the "Access external network" box and set the

priority to a lower value (higher number) than the one configured on VIOS 1.

· Remove SEA on VIOS 2 as it uses a different PVID than SEA on VIOS 1:

$ chdev -dev en2 -attr state=detach

$ rmdev -dev en2

en2 deleted

$ rmdev -dev et2

et2 deleted

$ rmdev -dev ent2

ent2 deleted

The virtual Ethernet adapter used with this SEA is of no use for SEA Failover and

may be removed.

· Create SEA with Failover feature on VIOS 2

$ mkvdev -sea ent0 -vadapter ent4 -default ent4 -defaultid 2 -attr

ha_mode=auto ctl_chan=ent3

ent2 Available

en2

et2

· The interface over the newly-configured SEA on VIOS 2 may now be

configured with an IP address in order to have network access to VIOS 2 again.

The client LPAR will run on the primary channel, even if VIOS 1 is down, as a

failover would now take place at the VIOS/SEA level. The NIB adapter may be

removed on the client LPAR at any time now.

Cases Initiating Failover:

· When the standby SEA detects that the keep-alive (heartbeat) messages are no

longer received over the control channel

· When the active SEA detects that a loss of physical link is reported by

physical Ethernet adapter’s device driver

· On VIOS with primary adapter, when ha_mode is manually set to standby

· When the active SEA detects it is no longer receiving replies from the IP

address it is pinging (if the “Internet Address to Ping” feature is enabled).

Test Cases for SEA Failover:

Here are the five scenarios where SEA Failover can be tested and should work:

· Manual Failover

ü Set ha_mode to standby on primary with chdev command:

chdev -dev <SEA device> -attr ha_mode=standby

ü Reset it back to auto and the SEA should fail back to the

primary:

chdev -dev <SEA device> -attr ha_mode=auto

ü

· Primary VIOS shutdown

ü Reboot the VIOS on the primary for fail over to backup

adapter

ü When the primary VIOS is up again, should fail back to the

primary adapter.

· Primary VIOS error

ü Deactivate primary VIOS from the HMC for fail over to

backup adapter

ü Activate the primary VIOS for the fail back to the primary

again

· Physical link failure

ü unplug the link of the physical adapter on primary for fail over

to the backup adapter

ü and replug of physical link of physical Ethernet adapter on

primary for the fail back to the primary

· Reverse boot sequence

ü Shut down both the VIO servers

ü Reboot the VIOS with backup SEA and the adapter becomes

active

ü Reboot the VIOS with primary SEA and fail back happens to

the primary SEA

Advantages of using SEA Failover over Network Interface Backup (NIB) on the client:

Network Interface Backup was used before the new failover feature was available -

pre-FixPack 7. This is implemented on the client for high availability in case one

VIOS went down. This involved at least a Shared Ethernet Adapter configured on

each of the two VIO servers with a different PVID on the trunk adapter on each of

them. A control channel is not required to be configured on the VIO servers. On the

client partition, a NIB is configured using two virtual Ethernet adapters, one with

PVID from the SEA on the VIOS1 and a second one with the PVID from the SEA on

the VIOS 2.

· In NIB, the client partition needs 2 virtual Ethernet adapters, implemented on the

client. In SEA Failover, only one virtual Ethernet adapter needs to be configured

on the client.

· The client partition configuration is simpler with SEA Failover: there is no logic

implemented for failover in the client as opposed to NIB.

· Redundancy and High Availability-related functions implemented on the VIOS

itself in SEA Failover.

· SEA Failover can be used with IEEE 802.1Q VLAN-tagging, while NIB cannot

be.

Troubleshooting:

· mkvdev to create SEA in ha_mode needs to be run in single step on the second

VIOS, not in two steps

· Please check that an IP address is not configured on either the virtual or the

physical Ethernet adapters that you are going to use for the SEA, otherwise the

configuration will fail.

· If you have multiple SEAs configured on each of the VIO servers, then for each

SEA pair, you need to configure a separate control channel with a unique PVID

on the system.

· If loading a new “seadd” driver, remove and recreate all the SEA adapters on the

VIOS to have the kernel extension active, a reboot is not required.

· The cable pull on the primary SEA adapter fails over to backup adapter but the

cable plug in back does not have the fail back to the primary if the SEA’s physical

adapter is an EtherChannel device --- there is an APAR already assigned to the

problem: IY82579.