Local failover not working

[James Langill]

Anyone every run into this problem? We have two internal lan cards and when testing MC Service Guard trying to have it do an internal failover from one lan card to another it will not work. If we pull the cable on the currently working Lan card it should just failover internally to the backup lan card but instead it failed over to the other machine. I would appreciate any help in this area as we are very unexperienced with MC/Service Guard.

Thanks,

Jim Langill

[James Langill]

No it is a completely seperate LAN card. Any suggestions?

[Andy Monks · Accepted solution]

It's not the same lan your running the heartbeat over is it?

If so, it's to be expected.

[Alan Riggs]

My understanding is that local lan failover is set automatically by the configuration daemon so long as your standby card is bridged to the same subnet as your primary and is in an unconfigured state. Does this describe your setup?

If so, what do the logfiles show for the period of the failover?

[Antoanetta Naghiu]

If in the same subnet, is not supported by HP.
And, if you really need help, answer to the questions above. All of them are valid questions....

[melvyn burnard]

ok, first few questions are:
what OS version?
what MC/SG version?
what type of lans?
are the lans bridged together? what does a cmquerycl -v -n nodea -n nodeb -C /tmp/testasciifile show ? does it show the lans as bridged?
have you set up a package yet with a monitored subnet?

If this is MC/SG 10.10 or above, run cmscancl and post the output file as an attachment.

[Gino Castoldi]

Hi,
I work with Jim Langill and I have some information that you have requested:

MC/SG version A.11.09 HP-UX 11.0 L2000 server
As for what lans are bridged together, a cmquerycl command shows the primary
LAN card as bridged(lan0), but it does not show the failover LAN card(lan2).

Note: LAN card 2(lan2) is not configured.

I also attached output from the cmscancl command.

Any help on this issue is appreciated. Thanks Gino.

[melvyn burnard]

according to your MC/SG binary file, you have no standby lan card configured for the
package subnet: 198.80.190.0

Therefore you will see the switching of the package.
I recoomend you:
a: check your ascii file
b: re-apply your configuration to correct the binary file
c: test the configuration

I would also recommend you have the MC/SG patch installed, if not already done: PHSS_21425

HTH

[Gino Castoldi]

Hi,
Melvyn, thanks for the information. After I ran a cmviewcl -v command:

CLUSTER STATUS
vpcluster up

NODE STATUS STATE
vpos1 up running

Network_Parameters:
INTERFACE STATUS PATH NAME
PRIMARY up 0/0/0/0 lan0
PRIMARY up 0/3/0/0 lan1

It shows that there is no Standby lan card configured as you said.

My question is this, how do you configure the package for a Standby lan card?

Thanks Gino

[melvyn burnard]

Well you need to first check that lan2 is acceptable as a backup lan for lan0 using the cmquerycl command. If this is ok, you need to edit or recreate your cluster ascii file and then re-apply the binary to the cluster.
Once you have done that, you need to restart the cluster and test the lan failover scenario
Please note you do NOT configure the package for the standby lan, but the cluster as a whole.

May I recommend reading the Managing MC/ServiceGuard manual at :
http://www.docs.hp.com/hpux/ha/#doc
Also you may want to take a look at the following course:

http://www.hp.com/education/courses/h6487s.html

or the newer one:

http://www.hp.com/education/courses/h4310s.html

This 3-day course provides system administrators who
manage existing MC/ServiceGuard clusters with the
knowledge and skills needed to extend their clusters and
increase the functionality of their packages. Topics include
growing the cluster to 4 nodes, using PRM to manage
package performance in failover situations and using EMS
to manage package dependencies. Successful completion
of HP MC/ServiceGuard (H6487S) is strongly
recommended before taking this course.

[Deshpande Prashant]

Hi
You should have entries like following in your cluster config file in /etc/cmcluster
----------
NODE_NAME
NETWORK_INTERFACE
HEARTBEAT_IP aa.bb.cc.dd
NETWORK_INTERFACE
NETWORK_INTERFACE
----------------

So here if lanx fails the same ip is bound to lany. And if both fails it binds to lanz.

Thanks.
Prashant

[Gino Castoldi]

Hi,

Thank you for the answers.

We coded in the parameters and we created a new file and we also tested
failover between the primary and standby lan cards and it worked fine!

Thank you again Gino and Jim.

I