CVM and Service guard heartbeat

[Vincent Abad]

Hi,

We are using VxVM + CVM 3.5 for running Oracle
9i RAC.
We have configured stanby lan interface under
heartbeat IP subnet(only one supported!!), and it's works fine in case of cable unplug.
But when the entire subnet fall down (switch failure) only one node survive, the other TOC and reboot.
I have 2 questions:

-Does the only work around is to configure a stanby LAN(switch).
-Does someone heard that Veritas plan to support multiple hearthbeat soon?.

Thanks,

--Vincent

[Artyom Voronchihin]

Hello.
You can use serial line (only for two nodes cluster) or multiple subnets for redundant heartbeat. One of the subnets is using as dedicated heartbeat LAN and other (say common data LAN) for redundant heartbeat. Subnets must be placed in different LAN segments to prevent failure if one of the segments goes down. Also dedicated heartbeat LAN can be used for data traffic between nodes.

[Michael Steele_2 · Accepted solution]

The heartbeat signal goes away during a NIC or switch failure causing the other node to be unable to obtain a cluster lock and reform properly. So your cluster design has to include redundant hubs or switches to account for the loss of signal when a one NIC or one switch fails. You need redundant switches.

Without the electrical signals sent out from a switch or hub the node believes the LAN is down and won't form.

[Michael Steele_2]

Let me add that the cluster relys upon the linkloop MAC address test and detects qualifying heartbeat LANs with the cmquerycl.