LAN failure within ServiceGuard Cluster

[KPS]

Hi,

Please help.. We're running ServiceGuard 11.17 on an rx8640. OS is 11.23 (ia64)

On our Primary Node, we lost connectivity to our primary NIC interface (lan1) 2 days ago and it failed over to our Stand-By (lan3). We have had our Network folks look at all that has to do with the switchport and all checks out fine. We think this was may have been a temporary inadvertent cable pull or something of the like. Just the same all looks to be back according to our Network folks and the switchport is all set and checking out fine with no errors. It is still setup on the same VLAN and Subnet. I check lanadmin and query the lan1 interface and it shows UP/UP. Linkloops from my Stand-By interface which is the one running the connectivity now in the cluster back to lan1 fail.

cmviewcl -v shows lan1 as still in a "DOWN" state. If that NIC starts talking to the network again, will ServiceGuard just show it back as "UP"?

We've done tests with the cable that goes to the NIC and unplugging it and plugging it into a laptop. We assign the laptop an IP in the same subnet as the server and ping the Broadcast Address and that works correctly and all IP's on that subnet answer. We can also ping the Gateway over that NIC.

Linkloop works if I run it against the lan1 interface and have it linkloop back to it's MAC.

Any ideas here would be greatly appreciated gurus!!!!!

Thanks,
KPS

[Ivan Krastev]

There are few possible reasons for fail:
- check speed/duplex settings;
- check VLAN assignment;
- broken NIC card;


regards,
ivan

[KPS]

- Speed and Duplex settings have been checked and no changes have occurred and they are correct.

- VLAN settings and config have been verified and all of that is unchanged and looks good.

- Broken NIC is where we are looking at presently... This is a Dual Gig-E card as well as Dual FC. What are the chances of a single port of the 2 Gig-E going south on us?

[melvyn burnard · Accepted solution]

One test to do is use the cmscancl command to check everything. This also does linkloop tests. Run it on both nodes to check completely. If a linkloop is failing, there is a physical issue with the connection, be it switch, cable or NIC.
Try also using lanadmin to reset the NIC.

If teh link were to come back top "life" Serviceguard should see this provided there is carrier, traffic etc.
One thing to also check is the patch levels of these servers.
And yes a single NIC port can fail on a Multi NIC card.

[KPS]

THanks for the suggestions and feedback on this problem. I will try some of these tests and suggestions the next chance we get.

/KPS