Forum Discussion

mkruer's avatar
mkruer
Level 4
12 years ago

Best practices for failover without heartbeat

I dont think there is a solution to this, becuse it would violate HA, but I need to ask the question.

Prerequisite: HA Setup is up and running.

Steps:
1). Unplug the heartbeat cable. (eth1)
2). Unplug/failover the primary system.

Secondary system does not come up without manually enabling the autodisable on
the faulted primary.

They want the secondary system to come up. Automaticly.

Is this possable? is it recomemned?

Best Practice
business continuity
Cluster Server
Configuring
Monitoring
  • Marianne's avatar
    Marianne
    Level 6
    12 years ago

    They want the secondary system to come up. Automaticly.

    WHY?? 
    What if primary is up and running, still having all clustered applications up and running?
    Have you ever seen 'split-brain'? Not pretty.
    Data corruption that can only be fixed by restoring all data.

    Autodisable is a protection mechanism. Understand that it is there to protect your data.

    Step 1 suggests that cluster has one heartbeat only. VCS needs two heartbeats.
    Please read up in VCS Admin Guide : 
    About communications, membership, and data protection in the cluster

    See these topics:

     

    ■ About data protection
    ■ About I/O fencing configuration files
    ■ Examples of VCS operation with I/O fencing
    ■ About cluster membership and data protection without I/O fencing
    ■ Examples of VCS operation without I/O fencing
    ■ Summary of best practices for cluster communications

    The solution here is to follow the 'best practices for cluster communications', e.g:

     

    ■ Configure multiple independent communication network links between cluster systems.
    Networks should not have a single point of failure, such as a shared switch or
    Ethernet card.
    ■ Low-priority LLT links in clusters with or without I/O fencing is recommended.
    In clusters without I/O fencing, this is critical.
    ...
    ■ Symantec recommends at least three coordination points to configure I/O
    fencing. You can use coordinator disks, CP servers, or a combination of both.
    Select the smallest possible LUNs for use as coordinator disks. No more than
    three coordinator disks are needed in any configuration.
     
    (More in the manual....)