Configuring split brain behavior-DRBD

BOoRFGOnZ

Configuring split brain behaviorSplit brain notification


DRBD invokes the split-brain handler, if        configured, at any time split brain is detected. To configure        this handler, add the following item to your resource        configuration:       
resource resource
  handlers {
    split-brain handler;
    ...
  }
  ...
}        handler may be any executable        present on the system.
Since DRBD version 8.2.6, the DRBD distribution contains a        split brain handler script that installs as        /usr/lib/drbd/notify-split-brain.sh. It        simply sends a notification e-mail message to        a specified address. To configure the handler to send a message        to root@localhost (which is expected to be an email        address that forwards the notification to a real system        administrator), configure the split-brain handler as        follows:       
resource resource
  handlers {
    split-brain "/usr/lib/drbd/notify-split-brain.sh root";
    ...
  }
  ...
}
After you have made this modfication on a running resource        (and synchronized the configuration file between nodes), no        additional intervention is needed to enable the handler. DRBD will        simply invoke the newly-configured handler on the next occurrence        of split brain.

Automatic split brain recovery policies


In order to be able to enable and configure DRBD's automatic        split brain recovery policies, you must understand that DRBD        offers several configuration options for this purpose. DRBD        applies its split brain recovery procedures based on the number        of nodes in the Primary role at the time the split brain is        detected. To that end, DRBD examines the following keywords, all        found in the resource's net configuration        section:      

• after-sb-0pri. Split brain has just been detected, but at this time                the resource is not in the Primary role on any host. For                this option, DRBD understands the following keywords:               
  
  • disconnect. Do not recover automatically, simply invoke                        the split-brain handler script (if                        configured), drop the connection and continue in                        disconnected mode.
  • discard-younger-primary. Discard and roll back the modifications made                        on the host which assumed the Primary role                        last.
  • discard-least-changes. Discard and roll back the modifications on                        the host where fewer changes occurred.
  • discard-zero-changes. If there is any host on which no changes                        occurred at all, simply apply all modifications                        made on the other and continue.
    
  
  
• after-sb-1pri. Split brain has just been detected, and at this time                the resource is in the Primary role on one host. For                this option, DRBD understands the following keywords:               
  
  • disconnect. As with after-sb-0pri, simply                        invoke the split-brain handler                        script (if configured), drop the connection and                        continue in disconnected mode.
  • consensus. Apply the same recovery policies as                        specified in after-sb-0pri. If a                        split brain victim can be selected after                        applying these policies, automatically resolve.                        Otherwise, behave exactly as if                        disconnect were specified.
  • call-pri-lost-after-sb. Apply the recovery policies as specified in                        after-sb-0pri. If a split brain                        victim can be selected after applying these                        policies, invoke the                        pri-lost-after-sb handler on the                        victim node. This handler must be configured in                        the handlers section and is                        expected to forcibly remove the node from the                        cluster.
  • discard-secondary. Whichever host is currently in the Secondary                        role, make that host the split brain                        victim.
    
  
  
• after-sb-2pri. Split brain has just been detected, and at this              time the resource is in the Primary role on both hosts.              This option accepts the same keywords as              after-sb-1pri except              discard-secondary and              consensus.
  

	Note
	DRBD understands additional keywords for these three          options, which have been omitted here because they are very          rarely used. Refer to drbd.conf(5) for          details on split brain recovery keywords not discussed          here.

For example, a resource which serves as the block device      for a GFS or OCFS2 file system in dual-Primary mode may have      its recovery policy defined as follows:       
resource resource {
  handlers {
    split-brain "/usr/lib/drbd/notify-split-brain.sh root"
    ...
  }
  net {
    after-sb-0pri discard-zero-changes;
    after-sb-1pri discard-secondary;
    after-sb-2pri disconnect;
    ...
  }
  ...
}