ethernet lan card failover capability ?

[Sammy_2]

Hi,
We are in 2-node cluster mc serviceguard environment(HP-Ux 11.0 OS on HP9000/N4000). we have installed two ethernet 100BT cards on each node. Is it possible to failover to 2nd lan card(lan2) just in case 1st lan card fails (lan1) ? Do we also have assign the same IP to the second card ? In other words, the second card should bind to the package IP address once the first one fails. That is failing over to the 2nd card on the same node. How do you configure it ?
==============================
Something like I have shown below. The IP
address below is the package IP address
===========================
lan1:1 10.101.102.103
should bind to
lan2:1 10.101.102.103
====================

Thanks a lot in advance.

[Craig Rants]

HP has some software called Auto Port Aggregation. Costs about $1500 per node. I have researched using it but don't have it as of yet. I believe that you can bind up to 5 ports to one ip. The cool thing is that it can replicate the MAC address so if you failover your ITO or OV stuff doesn't go nuts because an ip has a different MAC.

Good Luck,
C

[Christopher McCray_1]

Just doing a cmquerycl -v -n node1 -n node2 -C cluster.conf should do it for you. I would first do a cmgetconf first to get a temporary file of the original configuration first so you can retain all the other values when you edit the cluster configuration file. There should be an entry in your cluster.conf for your standby lan card afterward.

Hope this helps

Chris

[Christopher McCray_1]

The answer to your quastion is, if I may risk being redundant, is yes. MC/Service Guard allows for a failover lan card and, subsequently a place for the package ips to bind to. You need a minimum of 2 lan cards for MCSG anyway.

Good luck
Chris

[Sanjay_6 · Accepted solution]

Hi Sam,

The lan failover will not take place if the network card is not configured in your cluster configuration.

Say you have two lan cards and both the cards are configured as primary and are in use. In case one card fails, it won't switch over to the other card.

Now that you have added two more network cards, you have to configure these network cards into your cluster configuration, so the SG recognised these cards as standby cards and then will failover to these cards in case on a primary card failover.

you don;t have to assigh fresh ip's to the new cards. In case of the failure of a primary lan card, SG will switch all traffic to the standby lan port and will assign the primary port ip to the standby port.

Hope this helps.

Regds

[Uday_S_Ankolekar]

Hi,

Some info on configuring lan card

http://docs.hp.com/hpux/onlinedocs/B3936-90024/B3936-90024.html

and HA FAQ..

http://us-support.external.hp.com/iv/bin/doc.pl/sid=cc4d5bf606752687a6/screen=ivHome/?NODEID=English_SHW::WW_SW_UX_RCFAQ_EN_E/Q1.2&WARP=1

-USA..

[Christopher McCray_1]

Let me first apologize for this taking three tries to answer all your questions; it has been a bad morning. You will need to do nothing configuration wise to your lan2 card. Once MCSG recognizes that it is a standby lan card, in the event of lan card failure, MCSG will automatically configure the standy lan with the primary's configuration and the packages will also bind to it. If the primary lan card recovers or is fixed, the configuration reverts back to the original lan1.

I hope I finally answered everything!!!

Chris

[Sanjay_6]

Hi Sam,

Take a look at the thread below to find out more about how the SG stores the lan info,

http://us-support2.external.hp.com/cki/bin/doc.pl/sid=b3e7c68d119c5315a4/screen=ckiSearchResults?mode=id&searchString=KBRC00001282

Hope this helps.

Regds

[fg_1]

As long as the second nic is defined as the failover lan card, the primary IP will be bound to that NIC card upon failover. You can support multiple IP's on the same NIC but they have to be in the same subnet for that to work.

Port aggregation & teaming software could work for you as well.

By the way don't forget to put heartbeat's on each segment that is defined on your cluster. It helps with keeping up the pings.

[Roger Baptiste]

hi,

All you need is a unused lan port (of the same type as the primary lan port of the cluster). This port should be unassigned to any address.
eg: netstat -in
lan20* 1500 none none 0 0

Add the entry to the cluster conf file:
NETWORK_INTERFACE lan20
(after node name entry)

and run cmmakepkg command.

HTH
raj