failednode(p_failed_nodes integer, p_backup_node integer, p_failed_node integer[])

8.63. failednode(p_failed_nodes integer, p_backup_node integer, p_failed_node integer[])

Function Properties

Language: PLPGSQL

Return Type: integer

Initiate failover from failed_node to backup_node. This function must be called on all nodes, and then waited for the restart of all node daemons.

declare
	v_row				record;
	v_row2				record;
	v_failed					boolean;
    v_restart_required          boolean;
begin
	
	-- ----
	-- Grab the central configuration lock
	-- ----
	lock table sl_config_lock;

	v_restart_required:=false;
	--
	-- any nodes other than the backup receiving
	-- ANY subscription from a failed node
	-- will now get that data from the backup node.
	update sl_subscribe set 
		   sub_provider=p_backup_node
		   where sub_provider=p_failed_node
		   and sub_receiver<>p_backup_node
		   and sub_receiver <> ALL (p_failed_nodes);
	if found then
	   v_restart_required:=true;
	end if;
	-- ----
	-- Terminate all connections of the failed node the hard way
	-- ----
	perform terminateNodeConnections(p_failed_node);

	-- Clear out the paths for the failed node.
	-- This ensures that *this* node won't be pulling data from
	-- the failed node even if it *does* become accessible

	update sl_path set pa_conninfo='<event pending>' WHERE
	   		  pa_server=p_failed_node
			  and pa_conninfo<>'<event pending>';

	if found then
	   v_restart_required:=true;
	end if;

	v_failed := exists (select 1 from sl_node 
		   where no_failed=true and no_id=p_failed_node);

    if not v_failed then
	   	
		update sl_node set no_failed=true where no_id = ANY (p_failed_nodes)
			   and no_failed=false;
		if found then
	   	   v_restart_required:=true;
		end if;
	end if;	

	if v_restart_required then
	  -- Rewrite sl_listen table
	  perform RebuildListenEntries();	   
	
	  -- ----
	  -- Make sure the node daemon will restart
 	  -- ----
	  notify "_schemadoc_Restart";
    end if;


	-- ----
	-- That is it - so far.
	-- ----
	return p_failed_node;
end;