README.online-recovery

/*
 * $Header$
 *
 * pgpool: a language independent connection pool server for PostgreSQL 
 * written by Tatsuo Ishii
 *
 * Copyright (c) 2003-2010	PgPool Global Development Group
 *
 * Permission to use, copy, modify, and distribute this software and
 * its documentation for any purpose and without fee is hereby
 * granted, provided that the above copyright notice appear in all
 * copies and that both that copyright notice and this permission
 * notice appear in supporting documentation, and that the name of the
 * author not be used in advertising or publicity pertaining to
 * distribution of the software without specific, written prior
 * permission. The author makes no representations about the
 * suitability of this software for any purpose.  It is provided "as
 * is" without express or implied warranty.
 *
 */

How does online recovery work?

pcp_recovery_node calls pcp_recovery_node() API

pcp_recovery_node(int node_id) does:

  pcp_recovery_node sends "O" to pcp_child process requesting online
  recovery

    pcp_child calls start_recovery() and waits for its completion

	  start_recovery() sets Req_info->kind to NODE_RECOVERY_REQUEST
	    start_recovery calls execute_checkpoint() execute CHECKPOINT
	    command
	    start_recovery calls exec_recovery
	      exec_recovery executes SELECT pgpool_recovery()
		  execute 1st stage command		

		/* start 2nd stage */
		sets InRecovery to 1
		wait_connection_closed() to wait for all frontend exits
	    calls execute_checkpoint() execute CHECKPOINT
	    calls exec_recovery to execute 2nd stage script
		calls exec_remote_start to start postmaster
		calls check_postmaster_started to wait for postmaster starts
		up
		set pcp_wakeup request to 0
		calls send_failback_request()
		waits for failback completion every 1 second until
		pcp_wakeup_request becomes non 0.
		set pcp_wakeup request to 0

	  start_recovery calls finish_recovery()
	    finish_recovery sets InRecovery to 0
		sends SIGUSR2 signal to parent

		  Upon receiving SIGUSR2 pgpool parent calls
		  kill_all_children() which sends SIGUSR2 to all pgpool
		  children

		  child pgpool processes wake up from pause()