Two node cluster affecting separate four node cluster

Discussion in 'Clustering' started by Izz Noland, Oct 23, 2006.

  1. Izz Noland

    Izz Noland Guest

    We have an application cluster that has been prod for a few weeks now. Once
    it went prod, we had a network failure in which this particular cluster was
    directly affected. The odd thing, a separate cluster (AD file server still
    in test) also went down. Different switch, no public heartbeat, et cetera.
    While in the process of bringing the application server up, we sent it a net
    stop clussvc command on the console. THIS COMMAND WAS RECOGNIZED AND
    PROCESSED BY THE SEPARATE FILE CLUSTER. Has anyone seen something like this.
    These are not our only two clusters, and we do not see behaviour like this
    on any of our other clusters. It appears to be a one-way communication from
    the app cluster to the file cluster. Any ideas?
    Thanks in advance,
    Izz
     
    Izz Noland, Oct 23, 2006
    #1
    1. Advertisements

  2. have never experienced that, seems a bit odd.

    when you say "THIS COMMAND WAS RECOGNIZED AND PROCESSED BY THE SEPARATE FILE
    CLUSTER"

    do you mean that all 4 nodes of your file cluster "processed" the "net stop"
    command ?

    so are you saying that the "cluster service" of 4 nodes in a seperate
    cluster did stop, although you were running a "net stop" command on the
    console of a different cluster / node?

    (sorry for the questions, I just want to get facts clear and precise)
     
    Edwin vMierlo, Oct 28, 2006
    #2
    1. Advertisements

  3. Izz Noland

    Izz Noland Guest

    Yes, that is what was happening. After a coupple weeks, I found the link
    between the two clusters. We setup teaming on the new AD File Cluster and I
    realized that Broadcom's LiveLink was pinging the other cluster to check for
    connectivity. If connectivity was lost, by convention, LiveLink assumed
    their was a network outage and shutdown the network connections (and thus the
    cluster itself) to all four nodes in the AD file cluster. The net stop
    clussvc on the separate cluster was followed by a reboot, so this was
    coincidental. My recommendation, LiveLink for any time of teaming is NOT a
    good idea. I removed LiveLink altogether and failover from one NIC to the
    other is just a fast, and does not affect the cluster resources on that
    particular node. Seems like a waste to me, and was quite a headache for a
    couple of weeks.
    Thanks for the response, it was not the best move on my part in the first
    place!
    Izz
     
    Izz Noland, Oct 30, 2006
    #3
    1. Advertisements

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.