Megatest

http-transportmod.scm at [1b388397ae]
Login

File http-transportmod.scm artifact 55e6935b48 part of check-in 1b388397ae


;;======================================================================
;; Copyright 2017, Matthew Welland.
;; 
;; This file is part of Megatest.
;; 
;;     Megatest is free software: you can redistribute it and/or modify
;;     it under the terms of the GNU General Public License as published by
;;     the Free Software Foundation, either version 3 of the License, or
;;     (at your option) any later version.
;; 
;;     Megatest is distributed in the hope that it will be useful,
;;     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/>.

;;======================================================================

(declare (unit http-transportmod))
(declare (uses debugprint))
(declare (uses commonmod))
(declare (uses configfmod))
(declare (uses mtargs))
(declare (uses mtver))
(declare (uses dbmod))
(declare (uses stml2))
(declare (uses portloggermod))
(declare (uses pkts))

(module http-transportmod
	*
	
(import scheme
	(prefix sqlite3 sqlite3:)
	chicken.base
	chicken.condition
	chicken.file
	chicken.file.posix
	chicken.format
	chicken.io
	chicken.pathname
	chicken.port
	chicken.pretty-print
	chicken.process
	chicken.process-context
	chicken.process-context.posix
	chicken.sort
	chicken.string
	chicken.time
	chicken.time.posix
	chicken.tcp
	
	(prefix base64 base64:)
	directory-utils
	matchable
	md5
	message-digest
	regex
	s11n
	srfi-1
	srfi-13
	srfi-18
	srfi-69
	stack
	system-information
	typed-records
	z3
	  spiffy
	  uri-common
	  intarweb
	  http-client
	  spiffy-request-vars
	  intarweb
	  spiffy-directory-listing

	(prefix mtargs args:)
	commonmod
	configfmod
	debugprint
	mtver
	dbmod
	stml2
	portloggermod
	pkts
	
	)

;; (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) 

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

;;======================================================================
;; S E R V E R
;; ======================================================================

;; NOTE: http-transport:launch is the entry point
;;          -> http-transport:run
;;             -> http-transport:try-start-server -> http-transport:try-start-server (until success)

(define (http-get-function fnkey)
  (hash-table-ref/default *http-functions* fnkey (lambda () "nothing here yet")))

(define (http-transport:run hostn)
  ;; Configurations for server
  (tcp-buffer-size 2048)
  (max-connections 2048) 
  (debug:print 2 *default-log-port* "Attempting to start the server ...")
  (let* ((db              #f) ;;        (open-db)) ;; we don't want the server to be opening and closing the db unnecesarily
	 (hostname        (get-host-name))
	 (ipaddrstr       (let ((ipstr (if (string=? "-" hostn)
					   ;; (string-intersperse (map number->string (u8vector->list (hostname->ip hostname))) ".")
					   (server:get-best-guess-address hostname)
					   #f)))
			    (if ipstr ipstr hostn))) ;; hostname))) 
	 (start-port      (portlogger:open-run-close portlogger:find-port))
	 (link-tree-path  (common:get-linktree))
	 (tmp-area        (common:get-db-tmp-area))
	 (start-file      (conc tmp-area "/.server-start")))
    (debug:print-info 0 *default-log-port* "portlogger recommended port: " start-port)
    ;; set some parameters for the server
    (root-path     (if link-tree-path 
		       link-tree-path
		       (current-directory))) ;; WARNING: SECURITY HOLE. FIX ASAP!
    (handle-directory spiffy-directory-listing)
    (handle-exception (lambda (exn chain)
			(signal (make-composite-condition
				 (make-property-condition 
				  'server
				  'message "server error")))))

    ;; http-transport:handle-directory) ;; simple-directory-handler)
    ;; Setup the web server and a /ctrl interface
    ;;
    (vhost-map `(((* any) . ,(lambda (continue)
			       ;; open the db on the first call 
				 ;; 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 ;; the $ is the request vars proc
				    body: ((http-get-function 'api:process-request) *dbstruct-db* $)
				    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-get-function 'http-transport:main-page))))
				  ((equal? (uri-path (request-uri (current-request))) 
					   '(/ "json_api"))
				   (send-response body: ((http-get-function 'http-transport:main-page))))
				  ((equal? (uri-path (request-uri (current-request))) 
					   '(/ "runs"))
				   (send-response body: ((http-get-function 'http-transport:main-page))))
				  ((equal? (uri-path (request-uri (current-request))) 
					   '(/ any))
				   (send-response body: "hey there!\n"
						  headers: '((content-type text/plain))))
				  ((equal? (uri-path (request-uri (current-request))) 
					   '(/ "hey"))
				   (send-response body: "hey there!\n" 
						  headers: '((content-type text/plain))))
                                  ((equal? (uri-path (request-uri (current-request))) 
					   '(/ "jquery3.1.0.js"))
				   (send-response body: ((http-get-function 'http-transport:show-jquery))
						  headers: '((content-type application/javascript))))
                                  ((equal? (uri-path (request-uri (current-request))) 
					   '(/ "test_log"))
				   (send-response body: ((http-get-function 'http-transport:html-test-log) $) 
						  headers: '((content-type text/HTML))))    
                                  ((equal? (uri-path (request-uri (current-request))) 
					   '(/ "dashboard"))
				   (send-response body: ((http-get-function 'http-transport:html-dboard) $)
						  headers: '((content-type text/HTML)))) 
				  (else (continue))))))))
    #;(handle-exceptions
	exn
      (debug:print 0 *default-log-port* "Failed to create file " start-file ", exn=" exn)
      (with-output-to-file start-file (lambda ()(print (current-process-id)))))
    (http-transport:try-start-server ipaddrstr start-port)))

;; This is recursively run by http-transport:run until sucessful
;;
(define (http-transport:try-start-server ipaddrstr portnum)
  (let ((config-hostname (configf:lookup *configdat* "server" "hostname"))
	(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 
;;======================================================================

;;======================================================================
;; C L I E N T S
;;======================================================================


(define (http-transport:get-time-to-cleanup)
  (let ((res #f))
    (mutex-lock! *http-mutex*)
    (set! res (> (current-seconds) *http-connections-next-cleanup*))
    (mutex-unlock! *http-mutex*)
    res))

(define (http-transport:inc-requests-count)
  (mutex-lock! *http-mutex*)
  (set! *http-requests-in-progress* (+ 1 *http-requests-in-progress*))
  ;; Use this opportunity to slow things down iff there are too many requests in flight
  (if (> *http-requests-in-progress* 5)
      (begin
	(debug:print-info 0 *default-log-port* "Whoa there buddy, ease up...")
	(thread-sleep! 1)))
  (mutex-unlock! *http-mutex*))

(define (http-transport:dec-requests-count proc) 
  (mutex-lock! *http-mutex*)
  (proc)
  (set! *http-requests-in-progress* (- *http-requests-in-progress* 1))
  (mutex-unlock! *http-mutex*))

(define (http-transport:dec-requests-count-and-close-all-connections)
  (set! *http-requests-in-progress* (- *http-requests-in-progress* 1))
  (let loop ((etime (+ (current-seconds) 5))) ;; give up in five seconds
    (if (> *http-requests-in-progress* 0)
	(if (> etime (current-seconds))
	    (begin
	      (thread-sleep! 0.052)
	      (loop etime))
	    (debug:print-error 0 *default-log-port* "requests still in progress after 5 seconds of waiting. I'm going to pass on cleaning up http connections"))
	(close-idle-connections!)))
  (set! *http-connections-next-cleanup* (+ (current-seconds) 10))
  (mutex-unlock! *http-mutex*))

(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*))
         (server-id   (if (vector? serverdat) 
                           (http-transport:server-dat-get-server-id serverdat)
                           (begin
			     (debug:print 0 *default-log-port* "FATAL ERROR: http-transport:client-api-send-receive called with no server info")
			     (exit 1)))))
       (debug:print-info 11 *default-log-port* "cmd=" cmd " fullurl=" fullurl " server-id=" server-id " current time:" (current-seconds)) 

       ;; 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 
       ;; process and return it.
       (let* ((send-recieve (lambda ()
			      (mutex-lock! *http-mutex*)
			      ;; (condition-case (with-input-from-request "http://localhost"; #f read-lines)
			      ;;					       ((exn http client-error) e (print e)))
			      (set! res (vector                ;;; DON'T FORGET - THIS IS THE CLIENT SIDE! NOTE: consider moving this to client.scm since we are only supporting http transport at this time.
					 success
					 (db:string->obj 
					  (handle-exceptions
					      exn
					      (let ((call-chain (get-call-chain))
						    (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 " key:" (or server-id "thekey"))
                                                      (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-idle-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)
          (vector-set! res 0 success)
	 (thread-terminate! th2)
	 (if (vector? res)
	     (if (vector-ref res 0) ;; this is the first flag or the second flag? 
                 (let* ((res-dat (vector-ref res 1)))
                    (if (and (string? res-dat) (string-contains res-dat "server-id mismatch"))
                     (signal (make-composite-condition
		          (make-property-condition 
		       'servermismatch
		       'message  (vector-ref res 1))))       
		      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-get-server-id     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 server-id) 
  (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-id)))
    server-dat))

;;======================================================================
;; NEW SERVER METHOD
;;======================================================================

(define *srvpktspec*
  `((server (host    . h)
	    (port    . p)
	    (servkey . k)
	    (pid     . i)
	    (ipaddr  . a)
	    (dbpath  . d))))

(define (register-server pkts-dir pkt-spec host port servkey ipaddr dbpath)
  (let* ((pkt-dat `((host    . ,host)
		    (port    . ,port)
		    (servkey . ,servkey)
		    (pid     . ,(current-process-id))
		    (ipaddr  . ,ipaddr)
		    (dbpath  . ,dbpath))))
    (write-alist->pkt
     pkts-dir
     pkt-dat
     pktspec: pkt-spec
     ptype: 'server)))

;; ya, fake it for now
;;
(define (register-server-in-db db-file)
  #t)

(define (get-pkts-dir)
  (assert *toppath* "ERROR: get-pkts-dir called without *toppath* set. Exiting.")
  (let* ((pdir (conc *toppath* "/.meta/srvpkts")))
     (if (file-exists? pdir)
	 pdir
	 (begin
	   (create-directory pdir #t)
	   pdir))))

;; given a pkts dir read 
;;
(define (get-all-server-pkts pktsdir-in pktspec)
  (let* ((pktsdir  (if (file-exists? pktsdir-in)
		       pktsdir-in
		       (begin
			 (create-directory pktsdir-in #t)
			 pktsdir-in)))
	 (all-pkt-files (glob (conc pktsdir "/*.pkt"))))
    (map (lambda (pkt-file)
	   (read-pkt->alist pkt-file pktspec: pktspec))
	 all-pkt-files)))

(define (server-address srv-pkt)
  (conc (alist-ref 'host srv-pkt) ":"
	(alist-ref 'port srv-pkt)))
	
(define (server-ready? server-address)
  ;; ping the server and ask it
  ;; if it ready
  #f)

;; from the pkts return servers associated with dbpath
;; NOTE: Only one can be alive - have to check on each
;;       in the list of pkts returned
;;
(define (get-viable-servers serv-pkts dbpath)
  (let loop ((tail serv-pkts)
	     (res  '()))
    (if (null? tail)
	res ;; NOTE: sort by age so oldest is considered first
	(let* ((spkt (car tail)))
	  (loop (cdr tail)
		(if (equal? dbpath (alist-ref 'dbpath spkt))
		    (cons spkt res)
		    res))))))

;; from viable servers get one that is alive and ready
;;
(define (get-the-server serv-pkts dbpath)
  (let loop ((tail serv-pkts))
    (if (null? tail)
	#f
	(let* ((spkt (car tail))
	       (addr (server-address spkt)))
	  (if (server-ready? addr)
	      spkt
	      (loop (cdr tail)))))))

;; am I the "first" in line server? I.e. my D card is smallest
;; use Z card as tie breaker
;;
(define (get-best-candidate serv-pkts dbpath)
  (if (null? serv-pkts)
      #f
      (let loop ((tail serv-pkts)
		 (best  (car serv-pkts)))
	(if (null? tail)
	    best
	    (let* ((candidate (car tail))
		   (candidate-bd (string->number (alist-ref 'D candidate)))
		   (best-bd      (string->number (alist-ref 'D best)))
		   ;; bigger number is younger
		   (candidate-z  (alist-ref 'Z candidate))
		   (best-z       (alist-ref 'Z best))
		   (new-best     (cond
				  ((> best-bd candidate-bd) ;; best is younger than candidate
				   candidate)
				  ((< best-bd candidate-bd) ;; candidate is younger than best
				   best)
				  (else
				   (if (string>=? best-z candidate-z)
				       best
				       candidate))))) ;; use Z card as tie breaker
	      (if (null? tail)
		  new-best
		  (loop (cdr tail) new-best)))))))
	  
	  

;;======================================================================
;; END NEW SERVER METHOD
;;======================================================================

;; 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* ((run-id            (let ((rid (args:get-arg "-run-id")))
			      (if rid
				  (string->number rid)
				  #f)))
	 (db-file           (db:run-id->path run-id))
	 (sdat              #f)
	 (tmp-area          (common:get-db-tmp-area))
	 (server-start-time (current-seconds))
	 (pkts-dir          (get-pkts-dir))
	 (server-key        (server:mk-signature))
	 (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, now attempting to lock in server")
				;; create a server pkt in *toppath*/.meta/srvpkts
				(register-server pkts-dir *srvpktspec* (get-host-name)
						 (cadr sdat) server-key (car sdat) db-file)

				;; now read pkts and see if we are a contender
				(let* ((all-pkts     (get-all-server-pkts pkts-dir *srvpktspec*))
				       (viables      (get-viable-servers all-pkts db-file))
				       (best-srv     (get-best-candidate viables db-file))
				       (best-srv-key (if best-srv (alist-ref 'servkey best-srv) #f)))
				  ;; am I the best-srv, compare server-keys to know
				  (if (and (equal? best-srv-key server-key)
					   (register-server-in-db db-file))
				      (if (db:get-iam-server-lock *dbstruct-db* run-id)
					  (debug:print 0 *default-log-port* "I'm the server!")
					  (bdat-time-to-exit-set! *bdat* #t))) ;; nope, we are not needed, exit when can do
				  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-log-file (args:get-arg "-log"))) ;; always set when we are a server

    (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 *dbstruct-db* )
	  (let ((watchdog (bdat-watchdog *bdat*)))
	    (debug:print 0 *default-log-port* "SERVER: dbprep")
	    (db:setup run-id) ;; sets *dbstruct-db* as side effect
	    (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.
	    (if watchdog
		(if (not (member (thread-state watchdog) '(ready running blocked sleeping dead)))
		    (begin
		      (debug:print-info 0 "Starting watchdog thread (in state "(thread-state watchdog)")")
		      (thread-start! watchdog)))
		(debug:print 0 *default-log-port* "ERROR: *watchdog* not setup, cannot start it."))))
      
      ;; 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))
	    (thread-sleep! rem-time)))
      
      (if (< count 1) ;; 3x3 = 9 secs aprox
	  (loop (+ count 1) 'running bad-sync-count (current-milliseconds)))
      
      ;; Check that iface and port have not changed (can happen if server port collides)
      (mutex-lock! *heartbeat-mutex*)
      (set! sdat *server-info*)
      (mutex-unlock! *heartbeat-mutex*)
      
      (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)
             (if (not *server-id*)
              (set! *server-id* (server:mk-signature)))
            (debug:print 0 *default-log-port* (current-seconds) (current-directory) (current-process-id) (argv))
	    (debug:print 0 *default-log-port* "SERVER STARTED: " iface ":" port " AT " (current-seconds) " server-id: " *server-id*)
	    (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
             (if (not *server-id*)
              (set! *server-id* (server:mk-signature)))
            (debug:print 0 *default-log-port* (current-seconds) (current-directory) (current-process-id) (argv))   
	    (debug:print 0 *default-log-port* "SERVER STARTED: " iface ":" port " AT " (current-seconds) " server-id: " *server-id*)
	    (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 (and server-log-file (not *server-overloaded*))
		      (set-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
    ;;(BB> "http-transport:server-shutdown called")
    (debug:print-info 0 *default-log-port* "Starting to shutdown the server. pid="(current-process-id))
    ;;
    ;; start_shutdown
    ;;
    (bdat-time-to-exit-set! *bdat* #t) ;; tell on-exit to be fast as we've already cleaned up
    (portlogger:open-run-close portlogger:set-port port "released")
    (thread-sleep! 1)

    ;; (debug:print-info 0 *default-log-port* "Max cached queries was    " *max-cache-size*)
    ;; (debug:print-info 0 *default-log-port* "Number of cached writes   " *number-of-writes*)
    ;; (debug:print-info 0 *default-log-port* "Average cached write time "
    ;; 		      (if (eq? *number-of-writes* 0)
    ;; 			  "n/a (no writes)"
    ;; 			  (/ *writes-total-delay*
    ;; 			     *number-of-writes*))
    ;; 		      " ms")
    ;; (debug:print-info 0 *default-log-port* "Number non-cached queries "  *number-non-write-queries*)
    ;; (debug:print-info 0 *default-log-port* "Average non-cached time   "
    ;; 		      (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)))

;; Call this to start the actual server
;;

;; all routes though here end in exit ...
;;
;; start_server? 
;;
(define (http-transport:launch)
  (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)
                                (let* ((serv-fname      (conc "server-" (current-process-id) "-" (get-host-name) ".log"))
                                       (full-serv-fname (conc *toppath* "/logs/" serv-fname))
                                       (new-serv-fname  (conc *toppath* "/logs/" "defunct-" serv-fname)))
                                  (debug:print 0 *default-log-port* msg)
                                  (if (common:file-exists? full-serv-fname)
                                      (system (conc "sleep 1;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)))))
    (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! 0.252) ;; 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 ()
;; 			     (thread-sleep! 1))
;; 			   "eat response"))
;; 	 (th2 (make-thread (lambda ()
;; 			     (debug:print-error 0 *default-log-port* "Received ^C, attempting clean exit. Please be patient and wait a few seconds before hitting ^C again.")
;; 			     (thread-sleep! 3) ;; give the flush three seconds to do it's stuff
;; 			     (debug:print 0 *default-log-port* "       Done.")
;; 			     (exit 4))
;; 			   "exit on ^C timer")))
;;      (thread-start! th2)
;;      (thread-start! th1)
;;      (thread-join! th2))))

	    
;; Generate a unique signature for this server
(define (server:mk-signature)
  (message-digest-string (md5-primitive) 
			 (with-output-to-string
			   (lambda ()
			     (write (list (current-directory)
                                          (current-process-id)
					  (argv)))))))

(define (server:get-client-signature) 
  (if *my-client-signature* *my-client-signature*
      (let ((sig (server:mk-signature)))
        (set! *my-client-signature* sig)
        *my-client-signature*)))

;; run ping in separate process, safest way in some cases
;;
(define (server:ping-server ifaceport)
  (with-input-from-pipe 
   (conc (common:get-megatest-exe) " -ping " ifaceport)
   (lambda ()
     (let loop ((inl (read-line))
		(res "NOREPLY"))
       (if (eof-object? inl)
	   (case (string->symbol res)
	     ((NOREPLY)  #f)
	     ((LOGIN_OK) #t)
	     (else       #f))
	   (loop (read-line) inl))))))


;;======================================================================
;; S E R V E R
;;======================================================================

;; Call this to start the actual server
;;

;; all routes though here end in exit ...
;;
;; start_server
;;
(define (server:launch run-id transport-type)
    (http-transport:launch))

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

;; NOT USED (well, ok, reference in rpc-transport but otherwise not used).
;;
(define (server:login toppath)
  (lambda (toppath)
    (set! *db-last-access* (current-seconds)) ;; might not be needed.
    (if (equal? *toppath* toppath)
	#t
	#f)))

;; (define server:sync-lock-token "SERVER_SYNC_LOCK")
;; (define (server:release-sync-lock)
;;   (db:no-sync-del! *no-sync-db* server:sync-lock-token))
;; (define (server:have-sync-lock?)
;;   (let* ((have-lock-pair (db:no-sync-get-lock *no-sync-db* server:sync-lock-token))
;;          (have-lock?     (car have-lock-pair))
;;          (lock-time      (cdr have-lock-pair))
;;          (lock-age       (- (current-seconds) lock-time)))
;;     (cond
;;      (have-lock? #t)
;;      ((>lock-age
;;        (* 3 (configf:lookup-number *configdat* "server" "minimum-intersync-delay" default: 180)))
;;       (server:release-sync-lock)
;;       (server:have-sync-lock?))
;;      (else #f))))


)