topic Re: No failover when both data NICs fail in Operating System - HP-UX

No failover when both data NICs fail

Rudy Williams — Thu, 07 Apr 2011 14:50:06 GMT

Hello--

We have a cluster of two rx2800 i2 servers running HP-UX 11.31 and SG 11.19.

One NIC is on the data subnet, one NIC is standby and two NICs are attached to stand-alone small switches for heart beat. IN ServiceGuard, all NICs are configured as heartbeat.

When the primary data NIC fails, the standby NIC picks up just fine. When both data NICs fail, the package does NOT fail over to the adoptive node even though SG recognized both lan0 and lan2 have failed. When this occurs, each node does not know the state of the other. The output of cmviewcl shows "unknown" for the state of the other node.

I have been working with ServiceGuard for 13 years and have not seen this before. This is my first time working with version 11.19 of ServiceGuard.

I would appreciate any thoughts on the issue.

Re: No failover when both data NICs fail

melvyn burnard — Thu, 07 Apr 2011 15:02:08 GMT

is the package configured to monitor the subnet?
If not, then it will not failover.
attach the package configuration file if you want it to be reviewed

Re: No failover when both data NICs fail

Rudy Williams — Thu, 07 Apr 2011 15:34:59 GMT

Melvin--

No, we aren't monitoring the subnet in the package config file. I will edit, test and post an update. Thanks.

Re: No failover when both data NICs fail

melvyn burnard — Thu, 07 Apr 2011 17:21:03 GMT

for the package manager to move a package on subnet failure, you need to have the package monitoring the subnet. if using Legacy package format it will be in the package config file
If using Modular package format, it will also be in the package config file if you included the module

Re: No failover when both data NICs fail

Rudy Williams — Thu, 07 Apr 2011 18:48:44 GMT

Lacking a subnet monitor in the package configuration was exactly the problem. I added it and tested. The package is now failing as it ought to.

Thank you, sir!