Megatest

;;     but WITHOUT ANY WARRANTY; without even the implied warranty of
;;     MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
;;     GNU General Public License for more details.
;; 
;;     You should have received a copy of the GNU General Public License
;;     along with Megatest.  If not, see <http://www.gnu.org/licenses/>.

(require-extension (srfi 18) extras tcp s11n)


(use  srfi-1 posix regex regex-case srfi-69 hostinfo md5 message-digest posix-extras)

(use spiffy uri-common intarweb http-client spiffy-request-vars intarweb spiffy-directory-listing)

;; Configurations for server
(tcp-buffer-size 2048)
(max-connections 2048) 

(declare (unit http-transport))

(declare (uses common))

(declare (uses db))
(declare (uses tests))
(declare (uses tasks)) ;; tasks are where stuff is maintained about what is running.
(declare (uses server))
;; (declare (uses daemon))
(declare (uses portlogger))
(declare (uses rmt))























(include "common_records.scm")
(include "db_records.scm")
(include "js-path.scm")


(use stml2)

(define (http-transport:make-server-url hostport)
  (if (not hostport)
      #f
      (conc "http://" (car hostport) ":" (cadr hostport))))

(define *server-loop-heart-beat* (current-seconds))

				 ;; This is were we set up the database connections
			       (let* (($   (request-vars source: 'both))
				      (dat ($ 'dat))
				      (res #f))
				 (cond
				  ((equal? (uri-path (request-uri (current-request)))
					   '(/ "api"))
				   (send-response body:    (api:process-request *dbstruct-db* $) ;; the $ is the request vars proc
						  headers: '((content-type text/plain)))
				   (mutex-lock! *heartbeat-mutex*)
				   (set! *db-last-access* (current-seconds))
				   (mutex-unlock! *heartbeat-mutex*))
				  ((equal? (uri-path (request-uri (current-request))) 
					   '(/ ""))
				   (send-response body: (http-transport:main-page)))

	(config-use-proxy (equal? (configf:lookup *configdat* "client" "use-http_proxy") "yes")))
    (if (not config-use-proxy)
	(determine-proxy (constantly #f)))
    (debug:print-info 0 *default-log-port* "http-transport:try-start-server time=" (seconds->time-string (current-seconds)) " ipaddrsstr=" ipaddrstr " portnum=" portnum " config-hostname=" config-hostname)
    (handle-exceptions
	exn
	(begin
	  (print-error-message exn)
	  (if (< portnum 64000)
	      (begin 
		(debug:print 0 *default-log-port* "WARNING: attempt to start server failed. Trying again ...")
		(debug:print 0 *default-log-port* " message: " ((condition-property-accessor 'exn 'message) exn))
		(debug:print 5 *default-log-port* "exn=" (condition->list exn))
		(portlogger:open-run-close portlogger:set-failed portnum)
		(debug:print 0 *default-log-port* "WARNING: failed to start on portnum: " portnum ", trying next port")
		(thread-sleep! 0.1)
		
		;; get_next_port goes here
		(http-transport:try-start-server ipaddrstr
						 (portlogger:open-run-close portlogger:find-port)))
	      (begin
		(print "ERROR: Tried and tried but could not start the server"))))
      ;; any error in following steps will result in a retry
      (set! *server-info* (list ipaddrstr portnum))
      (debug:print 0 *default-log-port* "INFO: Trying to start server on " ipaddrstr ":" portnum)
      ;; This starts the spiffy server
      ;; NEED WAY TO SET IP TO #f TO BIND ALL
      ;; (start-server bind-address: ipaddrstr port: portnum)
      (if config-hostname ;; this is a hint to bind directly
	  (start-server port: portnum bind-address: (if (equal? config-hostname "-")
							ipaddrstr
							config-hostname))
	  (start-server port: portnum))
      (portlogger:open-run-close portlogger:set-port portnum "released")
      (debug:print 1 *default-log-port* "INFO: server has been stopped"))))

;;======================================================================
;; S E R V E R   U T I L I T I E S 
;;======================================================================

(define (http-transport:inc-requests-and-prep-to-close-all-connections)
  (mutex-lock! *http-mutex*)
  (set! *http-requests-in-progress* (+ 1 *http-requests-in-progress*)))

;; Send "cmd" with json payload "params" to serverdat and receive result
;;
(define (http-transport:client-api-send-receive run-id serverdat cmd params #!key (numretries 3)(area-dat #f))
  (let* ((fullurl    (if (vector? serverdat)
			 (http-transport:server-dat-get-api-req serverdat)
			 (begin
			   (debug:print 0 *default-log-port* "FATAL ERROR: http-transport:client-api-send-receive called with no server info")
			   (exit 1))))

	 (res        (vector #f "uninitialized"))
	 (success    #t)
	 (sparams    (db:obj->string params transport: 'http))
	 (runremote  (or area-dat *runremote*)))
       (debug:print-info 11 *default-log-port* "fullurl=" fullurl ", cmd=" cmd ", params=" params ", run-id=" run-id "\n")

       ;; set up the http-client here
       (max-retry-attempts 1)
       ;; consider all requests indempotent
       (retry-request? (lambda (request)
			 #f))
       ;; send the data and get the response
       ;; extract the needed info from the http data and 

						    (msg        ((condition-property-accessor 'exn 'message) exn)))
						(set! success #f)
                                                (if (debug:debug-mode 1)
                                                    (debug:print-info 0 *default-log-port* "couldn't talk to server, trying again ...")
                                                    (begin
                                                      (debug:print 0 *default-log-port* "WARNING: failure in with-input-from-request to " fullurl ".")
                                                      (debug:print 0 *default-log-port* " message: " msg ", exn=" exn)
                                                      (debug:print 0 *default-log-port* " cmd: " cmd " params: " params)
                                                      (debug:print 0 *default-log-port* " call-chain: " call-chain)))

                                                (if runremote
						    (remote-conndat-set! runremote #f))
						;; Killing associated server to allow clean retry.")
						;; (tasks:kill-server-run-id run-id)  ;; better to kill the server in the logic that called this routine?
						(mutex-unlock! *http-mutex*)
					     ;;; (signal (make-composite-condition
					     ;;;          (make-property-condition 'commfail 'message "failed to connect to server")))
					     ;;; "communications failed"
						(db:obj->string #f))
					    (with-input-from-request ;; was dat
					     fullurl 
					     (list (cons 'key (or *server-id* "thekey"))
						   (cons 'cmd cmd)
						   (cons 'params sparams))
					     read-string))
					  transport: 'http)
                                         0)) ;; added this speculatively
			      ;; Shouldn't this be a call to the managed call-all-connections stuff above?
			      (close-all-connections!)
			      (mutex-unlock! *http-mutex*)
			      ))
	      (time-out     (lambda ()
			      (thread-sleep! 45)
			      (debug:print 0 *default-log-port* "WARNING: send-receive took more than 45 seconds!!")
			      #f))
	      (th1 (make-thread send-recieve "with-input-from-request"))
	      (th2 (make-thread time-out     "time out")))
	 (thread-start! th1)
	 (thread-start! th2)
	 (thread-join! th1)

	 (thread-terminate! th2)
	 (debug:print-info 11 *default-log-port* "got res=" res)
	 (if (vector? res)
	     (if (vector-ref res 0) ;; this is the first flag or the second flag?






		 res ;; this is the *inner* vector? seriously? why?
                 (if (debug:debug-mode 11)
                     (let ((call-chain (get-call-chain))) ;; note: this code also called in nmsg-transport - consider consolidating it
                       (print-call-chain (current-error-port))
                       (debug:print-error 11 *default-log-port* "error above occured at server, res=" res) ;; " message: " ((condition-property-accessor 'exn 'message) exn))
                       (debug:print 11 *default-log-port* " server call chain:")
                       (pp (vector-ref res 1) (current-error-port))
                       (signal (vector-ref res 0)))
                     res))
	     (signal (make-composite-condition
		      (make-property-condition 
		       'timeout
		       'message "nmsg-transport:client-api-send-receive-raw timed out talking to server")))))))

;; careful closing of connections stored in *runremote*
;;
(define (http-transport:close-connections #!key (area-dat #f))
  (let* ((runremote  (or area-dat *runremote*))
	 (server-dat (if runremote
                         (remote-conndat runremote)
                         #f))) ;; (hash-table-ref/default *runremote* run-id #f)))
    (if (vector? server-dat)
	(let ((api-dat (http-transport:server-dat-get-api-uri server-dat)))
	  (handle-exceptions
	    exn
	    (begin
	      (print-call-chain *default-log-port*)
	      (debug:print-error 0 *default-log-port* " closing connection failed with error: " ((condition-property-accessor 'exn 'message) exn) ", exn=" exn))


	    (close-connection! api-dat)
            ;;(close-idle-connections!)
	    #t))
	#f)))


(define (make-http-transport:server-dat)(make-vector 6))
(define (http-transport:server-dat-get-iface         vec)    (vector-ref  vec 0))
(define (http-transport:server-dat-get-port          vec)    (vector-ref  vec 1))
(define (http-transport:server-dat-get-api-uri       vec)    (vector-ref  vec 2))

(define (http-transport:server-dat-get-api-url       vec)    (vector-ref  vec 3))
(define (http-transport:server-dat-get-api-req       vec)    (vector-ref  vec 4))
(define (http-transport:server-dat-get-last-access   vec)    (vector-ref  vec 5))
(define (http-transport:server-dat-get-socket        vec)    (vector-ref  vec 6))

(define (http-transport:server-dat-make-url vec)
  (if (and (http-transport:server-dat-get-iface vec)
	   (http-transport:server-dat-get-port  vec))
      (conc "http://" 
	    (http-transport:server-dat-get-iface vec)
	    ":"
	    (http-transport:server-dat-get-port  vec))
      #f))

(define (http-transport:server-dat-update-last-access vec)
  (if (vector? vec)
      (vector-set! vec 5 (current-seconds))
      (begin
	(print-call-chain (current-error-port))
	(debug:print-error 0 *default-log-port* "call to http-transport:server-dat-update-last-access with non-vector!!"))))

;;
;; connect
;;
(define (http-transport:client-connect iface port)
  (let* ((api-url      (conc "http://" iface ":" port "/api"))
	 (api-uri      (uri-reference (conc "http://" iface ":" port "/api")))
	 (api-req      (make-request method: 'POST uri: api-uri))
	 (server-dat   (vector iface port api-uri api-url api-req (current-seconds))))
    server-dat))

;; run http-transport:keep-running in a parallel thread to monitor that the db is being 
;; used and to shutdown after sometime if it is not.
;;
(define (http-transport:keep-running) 
  ;; if none running or if > 20 seconds since 
  ;; server last used then start shutdown
  ;; This thread waits for the server to come alive
  (debug:print-info 0 *default-log-port* "Starting the sync-back, keep alive thread in server")

  (let* ((sdat              #f)

	 (tmp-area          (common:get-db-tmp-area))
	 (started-file      (conc tmp-area "/.server-started"))
	 (server-start-time (current-seconds))
	 (server-info (let loop ((start-time (current-seconds))
				 (changed    #t)
				 (last-sdat  "not this"))
                        (begin ;; let ((sdat #f))
			  (thread-sleep! 0.01)
			  (debug:print-info 0 *default-log-port* "Waiting for server alive signature")
                          (mutex-lock! *heartbeat-mutex*)
                          (set! sdat *server-info*)
                          (mutex-unlock! *heartbeat-mutex*)
                          (if (and sdat
				   (not changed)
				   (> (- (current-seconds) start-time) 2))















			      (begin


				(debug:print-info 0 *default-log-port* "Received server alive signature")
                                (common:save-pkt `((action . alive)
                                                   (T      . server)
                                                   (pid    . ,(current-process-id))
                                                   (ipaddr . ,(car sdat))
                                                   (port   . ,(cadr sdat)))
                                                 *configdat* #t)
				sdat)
                              (begin
				(debug:print-info 0 *default-log-port* "Still waiting, last-sdat=" last-sdat)
                                (sleep 4)
				(if (> (- (current-seconds) start-time) 120) ;; been waiting for two minutes
				    (begin



				      (debug:print-error 0 *default-log-port* "transport appears to have died, exiting server")
                                      (common:save-pkt `((action . died)
                                                         (T      . server)
                                                         (pid    . ,(current-process-id))
                                                         (ipaddr . ,(car sdat))
                                                         (port   . ,(cadr sdat))
                                                         (msg    . "Transport died?"))
						       *configdat* #t)
				      (exit))

				    (loop start-time
					  (equal? sdat last-sdat)
					  sdat)))))))
	 (iface       (car server-info))
         (port        (cadr server-info))
         (last-access 0)
	 (server-timeout (server:expiration-timeout))
	 (server-going  #f)
	 (server-log-file (args:get-arg "-log"))) ;; always set when we are a server

    (handle-exceptions
	exn
      (debug:print 0 *default-log-port* "Failed to create " started-file ", exn=" exn)
      (with-output-to-file started-file (lambda ()(print (current-process-id)))))

    (let loop ((count         0)
	       (server-state 'available)
	       (bad-sync-count 0)
	       (start-time     (current-milliseconds)))

      ;; Use this opportunity to sync the tmp db to megatest.db
      (if (not server-going) ;; *dbstruct-db* 
	  (begin
	    (debug:print 0 *default-log-port* "SERVER: dbprep")
	    (set! *dbstruct-db*  (db:setup #t)) ;;  run-id))
	    (set! server-going #t)
	    (debug:print 0 *default-log-port* "SERVER: running, megatest version: " (common:get-full-version)) ;; NOTE: the server is NOT yet marked as running in the log. We do that in the keep-running routine.






	    (thread-start! *watchdog*)))
      
      ;; when things go wrong we don't want to be doing the various queries too often
      ;; so we strive to run this stuff only every four seconds or so.
      (let* ((sync-time (- (current-milliseconds) start-time))
	    (rem-time  (quotient (- 4000 sync-time) 1000)))
	(if (and (<= rem-time 4)
		 (>  rem-time 0))

      
      (if (not (equal? sdat (list iface port)))
	  (let ((new-iface (car sdat))
		(new-port  (cadr sdat)))
	    (debug:print-info 0 *default-log-port* "WARNING: interface changed, refreshing iface and port info")
	    (set! iface new-iface)
	    (set! port  new-port)


	    (debug:print 0 *default-log-port* "SERVER STARTED: " iface ":" port " AT " (current-seconds))
	    (flush-output *default-log-port*)))
      
      ;; Transfer *db-last-access* to last-access to use in checking that we are still alive
      (mutex-lock! *heartbeat-mutex*)
      (set! last-access *db-last-access*)
      (mutex-unlock! *heartbeat-mutex*)
      
      (if (common:low-noise-print 120 (conc "server running on " iface ":" port))
	  (begin



	    (debug:print 0 *default-log-port* "SERVER STARTED: " iface ":" port " AT " (current-seconds))
	    (flush-output *default-log-port*)))
      (if (common:low-noise-print 60 "dbstats")
	  (begin
	    (debug:print 0 *default-log-port* "Server stats:")
	    (db:print-current-query-stats)))
      (let* ((hrs-since-start  (/ (- (current-seconds) server-start-time) 3600)))
	(cond
         ((and *server-run*
	       (> (+ last-access server-timeout)
		  (current-seconds)))
          (if (common:low-noise-print 120 "server continuing")
              (debug:print-info 0 *default-log-port* "Server continuing, seconds since last db access: " (- (current-seconds) last-access))
	      (let ((curr-time (current-seconds)))
		(handle-exceptions
		    exn
		    (debug:print 0 *default-log-port* "ERROR: Failed to change timestamp on log file " server-log-file ". Are you out of space on that disk? exn=" exn)

		  (if (not *server-overloaded*)

		      (change-file-times server-log-file curr-time curr-time)))))









          (loop 0 server-state bad-sync-count (current-milliseconds)))
         (else
          (debug:print-info 0 *default-log-port* "Server timed out. seconds since last db access: " (- (current-seconds) last-access))
          (http-transport:server-shutdown port)))))))

(define (http-transport:server-shutdown port)
  (begin

    ;; 		      (if (eq? *number-non-write-queries* 0)
    ;; 			  "n/a (no queries)"
    ;; 			  (/ *total-non-write-delay* 
    ;; 			     *number-non-write-queries*))
    ;; 		      " ms")
    
    (db:print-current-query-stats)
    (common:save-pkt `((action . exit)
                       (T      . server)
                       (pid    . ,(current-process-id)))
                     *configdat* #t)



    (debug:print-info 0 *default-log-port* "Server shutdown complete. Exiting")
    (exit)))

;; all routes though here end in exit ...
;;
;; start_server? 
;;
(define (http-transport:launch)
  ;; check that a server start is in progress, pause or exit if so
  (let* ((tmp-area            (common:get-db-tmp-area))
	 (server-start        (conc tmp-area "/.server-start"))
	 (server-started      (conc tmp-area "/.server-started"))
	 (start-time          (common:lazy-modification-time server-start))
	 (started-time        (common:lazy-modification-time server-started))
	 (server-starting     (< start-time started-time)) ;; if start-time is less than started-time then a server is still starting
	 (start-time-old      (> (- (current-seconds) start-time) 5))
         (cleanup-proc        (lambda (msg) ;; would like to use (modulo (current-seconds) 60) instead of process-id to wrap filenames
                                (let* ((serv-fname      (conc "server-" (current-process-id) "-" (get-host-name) ".log"))
				       (new-fname       (conc "server-" (modulo (current-seconds) 60) "-" (get-host-name) ".log"))
                                       (full-serv-fname (conc *toppath* "/logs/" serv-fname))
                                       ;; (new-serv-fname  (conc *toppath* "/logs/" "defunct-" serv-fname))
				       (new-serv-fname  (conc *toppath* "/logs/" new-fname))
				       )
                                  (debug:print 0 *default-log-port* msg)
                                  (if (common:file-exists? full-serv-fname)
				      (with-output-to-pipe "at now + 10 minutes" (lambda ()
										   (print "mv -f " full-serv-fname " " new-serv-fname)))
                                      ;; (system (conc "sleep 10;mv -f " full-serv-fname " " new-serv-fname))
                                      (debug:print 0 *default-log-port* "INFO: cannot move " full-serv-fname " to " new-serv-fname))
                                  (exit)))))
    (if (and (not start-time-old) ;; last server start try was less than five seconds ago
	     (not server-starting))
	(begin
	  (cleanup-proc "NOT starting server, there is either a recently started server or a server in process of starting")
	  (exit)))
    ;; lets not even bother to start if there are already three or more server files ready to go
    (let* ((num-alive   (server:get-num-alive (server:get-list *toppath*))))
      (if (> num-alive 3)
          (begin
            (cleanup-proc (conc "ERROR: Aborting server start because there are already " num-alive " possible servers either running or starting up"))
            (exit))))
  (common:save-pkt `((action . start)
		     (T      . server)
		     (pid    . ,(current-process-id)))
		   *configdat* #t)
    (let* ((th2 (make-thread (lambda ()
                               (debug:print-info 0 *default-log-port* "Server run thread started")
                               (http-transport:run 
                                (if (args:get-arg "-server")
                                    (args:get-arg "-server")
                                    "-")
                                )) "Server run"))
           (th3 (make-thread (lambda ()
                               (debug:print-info 0 *default-log-port* "Server monitor thread started")
                               (http-transport:keep-running)
                               "Keep running"))))
      (thread-start! th2)
      (thread-sleep! 0.25) ;; give the server time to settle before starting the keep-running monitor.
      (thread-start! th3)
      (set! *didsomething* #t)
      (thread-join! th2)
      (exit))))

;; (define (http-transport:server-signal-handler signum)
;;   (signal-mask! signum)
;;   (handle-exceptions
;;    exn
;;    (debug:print 0 *default-log-port* " ... exiting ...")
;;    (let ((th1 (make-thread (lambda ()

    (conc "<table>"
	  (string-intersperse
	   (map (lambda (stat)
		  (conc "<tr><td>" (car stat) "</td><td>" (cadr stat) "</td></tr>"))
		stats)
	   " ")
	  "</table>")))