Serviceguard failover problems

[karthiknarayan]

we have installed and configured the HP MC Service Guard for clustering.
When we shutdown the system the failover of resource from system A to system B and from System B to System A are working fine. But failover doesn't happen if we stop the public interface card (i.e. LAN 0) or remote the patch cord from that interface.

Please Help me here

[Bernhard Mueller]

Hi,

if you are on MC/SG 11.16 please read the following in your cmclconf.ascii:

# Network Monitor Configuration Parameters.
# The NETWORK_FAILURE_DETECTION parameter determines how LAN card failures are detected.
# If set to INONLY_OR_INOUT, a LAN card will be considered down when its inbound
# message count stops increasing or when both inbound and outbound
# message counts stop increasing.
# If set to INOUT, both the inbound and outbound message counts must
# stop increasing before the card is considered down.
NETWORK_FAILURE_DETECTION INONLY_OR_INOUT

So setting to INONLY_OR_INOUT should initiate failover if you pull the patch cable into your production lan, if you have a standby card, you need to pull both. Plus your subnet needs to be monitored (set in the pkg.conf).

Regards,
Bernhad

[Fred Ruffet]

What is output for "cmviewcl -v" ?

Regards,

Fred

[Dietmar Konermann]

Karthiknarayan,

> if we stop the public interface card

What do you mean with "stop"? If you simply perform "ifconfig lan0 down", then indeed this is no failure SG is supposed to detect. The network sensor is designed to detect interface HW failures only... setting the IP state to down is not a HW failure.

> remote the patch cord from that interface

This indeed would cause the driver state of the interface to go DOWN which should be detected by SG. Don't you see messages like "lan0 down" from cmcld in syslog.log?

Have you configured the corresponding SUBNET to be monitored by the package? (Using the SUBNET statement in the package conf file?).

Best regards...
Dietmar.

[Geoff Wild]

Do you have a second (stand by) interface? If so - then package won't failover - only the interface - that way package stays up.

post a: cmgetconf

Rgds...Geoff

[Steven E. Protter]

That stop needs to be either a cable pull or switch shutdown. Since most of us admins don't like touching switches shared by othere, I recommend the cable pull test during a planned maintenance downtime.

While you are at it draw up some test commands to run to make sure packages fail over properly.

SEP

[Sridhar Bhaskarla]

Hi Karthiknarayan,

Can you post your configuration?. (your cluster ascii and package configuration files). Failover will not happen based on quite a few reasons.

1. If the IP is not part of serviceguard.
2. AUTORUN is disabled in package configuration.
3. If the subnet is not monitored in the package configuration. Search for "SUBNET" in the package configuration file.

-Sri