Forum Discussion

mokkan's avatar
mokkan
Level 6
12 years ago
Solved

main.cf update question on 2 node cluster

I have a quick quesiton regarding main.cf update.  Here are the steps I would like to do for 2 nodes cluster.

 

Node1:  hastop  -all

            cp   /etc/VRTSsvc/conf/confi/main.cf   main.cf.back

            cp  /etc/VRTSsvc/conf/confi/main.cf    /tmp/main.cf

            hastart -all

 

Is this right?  If so,  how broadcasting works?  When I start cluster at node1 would main.cf will copy to other nodes right? Is there a way node2 main.cf wil copy over to node1?  How does it determine which direction it needs to be copied?

Basics
business continuity
Cluster Server
UNIX
  • Gaurav_S's avatar
    Gaurav_S
    12 years ago

    Hi,

    I don't understand what you are trying to do here ... you are simply copying main.cf to some other directories ... (though the path you have mentioned is /etc/VRTSvcs/conf/config/main.cf  ..there is typo in your comment)

    Before you stop cluster using hastop -all, you should do a "haconf -dump -makero" in order to flush down the changes to main.cf file.

    Lets say you have executed hastop -all & now cluster is stopped. Which ever node you run hastart first will do a local build & upload the main.cf into kernel. when you run hastart on any other node now, it will do a remote build & the main.cf from first node which was already uploaded to kernel will travel via LLT links to this second node.

    There is no way that main.cf of second node (where you executed hastart later), its main.cf CAN NOT copy to first node (where hastart was executed first). I hope this answers your direction question also.

    G

  • Gaurav_S's avatar
    Gaurav_S
    Moderator
    12 years ago

    Hi,

    I don't understand what you are trying to do here ... you are simply copying main.cf to some other directories ... (though the path you have mentioned is /etc/VRTSvcs/conf/config/main.cf  ..there is typo in your comment)

    Before you stop cluster using hastop -all, you should do a "haconf -dump -makero" in order to flush down the changes to main.cf file.

    Lets say you have executed hastop -all & now cluster is stopped. Which ever node you run hastart first will do a local build & upload the main.cf into kernel. when you run hastart on any other node now, it will do a remote build & the main.cf from first node which was already uploaded to kernel will travel via LLT links to this second node.

    There is no way that main.cf of second node (where you executed hastart later), its main.cf CAN NOT copy to first node (where hastart was executed first). I hope this answers your direction question also.

    G

  • g_lee's avatar
    g_lee
    Level 6
    12 years ago

    Gaurav is correct - the first node to form the cluster does a remote build; nodes that subsequently join the get the config from the other running nodes already in the cluster.

    Reiterating the importance of haconf -dump -makero before the hastop -all to ensure all changes are saved prior to stopping the cluster.

    Also, note that hastop -all will offline/shutdown the groups. If application(s)/resources need to be left online while any changes are being made (not clear from steps provided), use hastop -all -force. Again, ensure the config is closed before issuing the stop (haconf -dump -makero).