Halt Script hangs on "ifalias delete"

[Rob van Eerd]

Hi,

We're using Remedy's Action Request System on an MC/ServiceGuard cluster with two K-360's running HP-UX 10.20.
Due to the way ARS works with the IP-Address assigned to the NIC, we have to use a bypass (thanks John Palmer for assisting me during setup).
During startup of the package:
- Assign the Package IP-addres to the NIC
- Assign the Box's IP-address as an alias to the NIC.

During halt of the package:
- Delete the alias (Box's IP-address) from the NIC
- Re-assign Box's IP address to the NIC.

Unfortunately our system hangs once the "ifalias lan1 delete ${BOX-IP-address}" is issued.
Only heartbeat-address replies on ping!
We have to switch off the power of the box to be able to continue!

I've attached the customer_defined_halt_cmnds for further information.

Can anyone assist? We're not able to switch in a proper way.

Oh yes. We're having this problem for some time now so don't ask me what changed. I'm affraid I won't be able to tell that.
So let's start on the current situation!

[John Palmer]

Hi Rob,

One thing that I note from your halt script is that you've added NFS at some point. I would ensure that all NFS mounts are removed before manipulating the LAN addresses.

Another thing - would you be able to upgrade to HP-UX 11? At that release, the problem with the source IP address goes away

Regards,
John.

[Rob van Eerd]

John,

Thanks for the answer.

NFS was only recently added. I've had the problem before that moment as well.
Although it's worth trying once and manually unmount the NFS-link first.

HP-UX 11.0 is an option but not for now. My users are waiting for the new features of EARS (which doesn't need HP-UX 11.0) so first an EARS upgrade and then an HP-UX upgrade.

What do you mean with: "At that release, the problem with the source IP address goes away".
Does this mean the fooling around with IP-addesses:
- assign the package-IP-addres to the NIC
- and the Box-Ip-addres as an alias
is not necessary anymore? That would mean I've got another option of upgrading to HP-UX first because we're not Highly Available at this moment.

Thanks,
Rob.

[John Palmer]

Hi Rob,

At 11.00 the aliasing of IP addresses has been totally modified. Instead of a second address on the same interface (which gives the source address problem), you get a new interface such as lan0:1 which removes the problem.

We haven't tested it with EARS yet even though we are on 11.00 but other access such as telnet indicates that the problem has gone away.

Regards,
John