Forum Discussion

Rajan_Iyer's avatar
Rajan_Iyer
Level 4
13 years ago

CFS Node Out Of Service

Hi,

I would like to know what the actions need to be taken to keep CFS Node Out of Service.

For example in 3 Node CFS Parallel Cluster one of the node is malfunctioning, in the same ref. I would like to troubleshoot the same in which I want to keep that node from CFS/CVM Cluster, so that application can see only two nodes in the cluster and will write the data from that working 2 nodes.

 

Once 3rd nodeis up I would like to take in CVM/CFS cluster.

 

Please let me know the command.

 

Thanks

Rajan

business continuity
Cluster Server
  • Gaurav_S's avatar
    Gaurav_S
    13 years ago

    ok ... then simply keep node of CVM by offlining CVM service group ... that should be enough..

3 Replies

Sort By
  • Gaurav_S's avatar
    Gaurav_S
    Moderator
    13 years ago

    Hi Rajan,

    What sort of maintenance you are performing on the 3rd node ?

    If you simply want to keep the node out of CVM cluster, just don't online the "cvm" group on the 3rd node, this will stop CVM cluster to form on the 3rd node.

    If you want to completely take out the node from cluster, then you need to stop the cluster services on the 3rd node. If the 3rd node is running, you can run "hagrp -offline cvm <node>" & then followed by hastop -local. Ofcourse you will need to offline all the other dependent service groups.

    If the node is not up, you need to form the cluster with 2 nodes only .. this can be accomplished by manually seeding the GAB.

     

    Let me know if this helps..

     

    Gaurav

  • Rajan_Iyer's avatar
    Rajan_Iyer
    Level 4
    13 years ago

    Hi gaurav,

    In my case I do not want that application will hit 3rd node for accessing parallel file system.

     

    I think removing from CVM membership it will help us.

     

    Thanks

    Rajan

  • Gaurav_S's avatar
    Gaurav_S
    Moderator
    13 years ago

    ok ... then simply keep node of CVM by offlining CVM service group ... that should be enough..