Make_net_recovery takes a long time

[George_231]

I am doing make_net_recovery on five Itanium B.11.23 servers and while three of them complete the make_net_recovery in 45-50 minutes, two of them take a lot of time, 16 hours or more. The servers have about the same size of the image created(3.3GB)
All of the servers are 6 vpars in 2 npars in one Superdome cabinet. The cabinet does not have a tape drive so I have to do the make_net_recovery on one of them acting as a server and the rest are clients.
The NIC cards configured in Ignite are used only for this purpose and application maintenance personel logging in. The client uses other front end NICs while storage backup is using additional cards for its purposes.
What could be the reason for this?

[DCE]

I have found that the compatability between LAN card speed settings and the switch settings are critical. Auto negotiation is not always the most optimal setting. You might want to verify that these settings are consistent across all of your Ignite NIC cards/switch connections

[V. Nyga]

Hi,

check your network statistics for errors or collisions - reset it and start your process again. If statistic is bad again, configure it like DCE mentioned.

Volkmar

[Bob Ingersoll]

As previously mentioned you likely have a mismatch between the Ignite client LAN interface and switch port that is causing errors. You can use the lanadmin command to check for errors:

lanadmin -g mibstats 0

Verify the interface configuration using SAM or lanadmin:

lanadmin -x 0

Replace 0 with the interface instance number displayed by lanscan.

Check for collisions.

The LAN interface and switch port used for backup/recovery should both be configured to autonegotiate. This is important because the install kernel used when you boot from the LAN to do a recovery configures the LAN interface to autonegotiate. If the switch port isn't configured to autonegotiate the recovery will fail or take hours to complete. So it's best to ensure both the interface and port are set to autonegotiate.

[George_231]

lanadmin -g mibstats showed collisions. Checked the lan configuration. It showed half duplex auto ON. Changed it to full duplex and now it is OK.
Thanks for all your help