Megatest

Diff
Login

Differences From Artifact [bedbeec6cc]:

To Artifact [9b05b6d402]:


23
24
25
26
27
28
29

30
31
32
33
34
35
36

(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))


(include "common_records.scm")
(include "db_records.scm")

(define (http-transport:make-server-url hostport)
  (if (not hostport)
      #f







>







23
24
25
26
27
28
29
30
31
32
33
34
35
36
37

(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))

(include "common_records.scm")
(include "db_records.scm")

(define (http-transport:make-server-url hostport)
  (if (not hostport)
      #f
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88

89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150


151
152
153
154
155
156




157
158
159
160

161





162
163
164
165
166
167
168
169
170
171
172
173


174




175
176
177
178
179
180
181
182
183
184
185



186
187
188
189
190
191
192
	   (set! res adr)))
     (vector->list (hostinfo-addresses (hostname->hostinfo hostname))))
    (string-intersperse 
     (map number->string
	  (u8vector->list
	   (if res res (hostname->ip hostname)))) ".")))

(define (http-transport:run hostn)
  (debug:print 2 "Attempting to start the server ...")
  (if (not *toppath*)
      (if (not (setup-for-run))
	  (begin
	    (debug:print 0 "ERROR: cannot find megatest.config, cannot start server, exiting")
	    (exit))))
  (let* (;; (iface           (if (string=? "-" hostn)
	 ;;        	      #f ;; (get-host-name) 
	 ;;        	      hostn))
	 (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    (if (and (args:get-arg "-port")
				 (string->number (args:get-arg "-port")))
			    (string->number (args:get-arg "-port"))
			    (if (and (config-lookup  *configdat* "server" "port")
				     (string->number (config-lookup  *configdat* "server" "port")))
				(string->number (config-lookup  *configdat* "server" "port"))
				(+ 5000 (random 1001)))))
	 (link-tree-path (config-lookup *configdat* "setup" "linktree")))
    (set! db *inmemdb*)

    (root-path     (if link-tree-path 
		       link-tree-path
		       (current-directory))) ;; WARNING: SECURITY HOLE. FIX ASAP!
    (handle-directory spiffy-directory-listing)
    ;; 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 body:    (api:process-request db $) ;; the $ is the request vars proc
						  headers: '((content-type text/plain)))
				   (mutex-lock! *heartbeat-mutex*)
				   (set! *last-db-access* (current-seconds))
				   (mutex-unlock! *heartbeat-mutex*))
				  ;; This is the /ctrl path where data is handed to the server and
				  ;; responses 
				  ((equal? (uri-path (request-uri (current-request)))
					   '(/ "ctrl"))
				   (let* ((packet (db:string->obj dat))
					  (qtype  (cdb:packet-get-qtype packet)))
				     (debug:print-info 12 "server=> received packet=" packet)
				     (if (not (member qtype '(sync ping)))
					 (begin
					   (mutex-lock! *heartbeat-mutex*)
					   (set! *last-db-access* (current-seconds))
					   (mutex-unlock! *heartbeat-mutex*)))
				     ;; (mutex-lock! *db:process-queue-mutex*) ;; trying a mutex
				     ;; (set! res (open-run-close db:process-queue-item open-db packet))
				     (set! res (db:process-queue-item db packet))
				     ;; (mutex-unlock! *db:process-queue-mutex*)
				     (debug:print-info 11 "Return value from db:process-queue-item is " res)
				     (send-response body: (conc "<head>ctrl data</head>\n<body>"
								res
								"</body>")
						    headers: '((content-type text/plain)))))
				  ((equal? (uri-path (request-uri (current-request))) 
					   '(/ ""))
				   (send-response body: (http-transport:main-page)))
				  ((equal? (uri-path (request-uri (current-request))) 
					   '(/ "runs"))
				   (send-response body: (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))))
				  (else (continue))))))))
    (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)


  (handle-exceptions
   exn
   (begin
     (print-error-message exn)
     (if (< portnum 9000)
	 (begin 




	   (debug:print 0 "WARNING: failed to start on portnum: " portnum ", trying next port")
	   (thread-sleep! 0.1)
	   ;; (open-run-close tasks:remove-server-records tasks:open-db)
	   (open-run-close tasks:server-delete tasks:open-db ipaddrstr portnum)

	   (http-transport:try-start-server ipaddrstr (+ portnum 1)))





	 (print "ERROR: Tried and tried but could not start the server")))
   ;; any error in following steps will result in a retry
   (set! *runremote* (list ipaddrstr portnum))
   ;; (open-run-close tasks:remove-server-records tasks:open-db)
   (open-run-close tasks:server-register 
		   tasks:open-db 
		   (current-process-id)
		   ipaddrstr portnum 0 'startup 'http)
   (debug:print 1 "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)


   (open-run-close tasks:server-delete tasks:open-db ipaddrstr portnum)




   (debug:print 1 "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-mutex* (make-mutex))




;; This next block all imported en-mass from the api branch
(define *http-requests-in-progress* 0)
(define *http-connections-next-cleanup* (current-seconds))

(define (http-transport:get-time-to-cleanup)
  (let ((res #f))







|

<
<
<
<
<
<
<
<
|






|
<
<
<
<
<
<
|
|
>
















|






|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|















|



|
>
>
|
|
|
|
|
|
>
>
>
>
|
|
|
<
>
|
>
>
>
>
>
|
|
|
<
|
|
|
|
|
|
|
|
>
>
|
>
>
>
>
|










>
>
>







57
58
59
60
61
62
63
64
65








66
67
68
69
70
71
72
73






74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153

154
155
156
157
158
159
160
161
162
163

164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
	   (set! res adr)))
     (vector->list (hostinfo-addresses (hostname->hostinfo hostname))))
    (string-intersperse 
     (map number->string
	  (u8vector->list
	   (if res res (hostname->ip hostname)))) ".")))

(define (http-transport:run hostn run-id server-id)
  (debug:print 2 "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  (configf:lookup *configdat* "setup" "linktree")))
    ;; (set! db *inmemdb*)
    (debug:print-info 0 "portlogger recommended port: " start-port)
    (root-path     (if link-tree-path 
		       link-tree-path
		       (current-directory))) ;; WARNING: SECURITY HOLE. FIX ASAP!
    (handle-directory spiffy-directory-listing)
    ;; 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 body:    (api:process-request *inmemdb* $) ;; the $ is the request vars proc
						  headers: '((content-type text/plain)))
				   (mutex-lock! *heartbeat-mutex*)
				   (set! *last-db-access* (current-seconds))
				   (mutex-unlock! *heartbeat-mutex*))
				  ;; This is the /ctrl path where data is handed to the server and
				  ;; responses 
				  ;; ((equal? (uri-path (request-uri (current-request)))
				  ;;          '(/ "ctrl"))
				  ;;  (let* ((packet (db:string->obj dat))
				  ;;         (qtype  (cdb:packet-get-qtype packet)))
				  ;;    (debug:print-info 12 "server=> received packet=" packet)
				  ;;    (if (not (member qtype '(sync ping)))
				  ;;        (begin
				  ;;          (mutex-lock! *heartbeat-mutex*)
				  ;;          (set! *last-db-access* (current-seconds))
				  ;;          (mutex-unlock! *heartbeat-mutex*)))
				  ;;    ;; (mutex-lock! *db:process-queue-mutex*) ;; trying a mutex
				  ;;    ;; (set! res (open-run-close db:process-queue-item open-db packet))
				  ;;    (set! res (db:process-queue-item db packet))
				  ;;    ;; (mutex-unlock! *db:process-queue-mutex*)
				  ;;    (debug:print-info 11 "Return value from db:process-queue-item is " res)
				  ;;    (send-response body: (conc "<head>ctrl data</head>\n<body>"
				  ;;       			res
				  ;;       			"</body>")
				  ;;       	    headers: '((content-type text/plain)))))
				  ((equal? (uri-path (request-uri (current-request))) 
					   '(/ ""))
				   (send-response body: (http-transport:main-page)))
				  ((equal? (uri-path (request-uri (current-request))) 
					   '(/ "runs"))
				   (send-response body: (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))))
				  (else (continue))))))))
    (http-transport:try-start-server run-id ipaddrstr start-port server-id)))

;; This is recursively run by http-transport:run until sucessful
;;
(define (http-transport:try-start-server run-id ipaddrstr portnum server-id)
  (let ((config-hostname (configf:lookup *configdat* "server" "hostname")))
    (debug:print-info 0 "http-transport:try-start-server run-id=" run-id " ipaddrsstr=" ipaddrstr " portnum=" portnum " server-id=" server-id " config-hostname=" config-hostname)
    (handle-exceptions
     exn
     (begin
       (print-error-message exn)
       (if (< portnum 64000)
	   (begin 
	     (debug:print 0 "WARNING: attempt to start server failed. Trying again ...")
	     (debug:print 0 " message: " ((condition-property-accessor 'exn 'message) exn))
	     (debug:print 0 "exn=" (condition->list exn))
	     (portlogger:open-run-close portlogger:set-failed portnum)
	     (debug:print 0 "WARNING: failed to start on portnum: " portnum ", trying next port")
	     (thread-sleep! 0.1)


	     ;; get_next_port goes here
	     (http-transport:try-start-server run-id
					      ipaddrstr
					      (portlogger:open-run-close portlogger:find-port)
					      server-id))
	   (begin
	     (tasks:server-force-clean-run-record (tasks:get-db) run-id ipaddrstr portnum " http-transport:try-start-server")
	     (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))

     (tasks:server-set-interface-port 
		     (tasks:get-db)
		     server-id 
		     ipaddrstr portnum)
     (debug:print 0 "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")
     (tasks:server-force-clean-run-record (tasks:get-db) run-id ipaddrstr portnum " http-transport:try-start-server")
     (debug:print 1 "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-mutex* (make-mutex))

;; NOTE: Large block of code from 32436b426188080f72fceb6894af541fbad9921e removed here
;;       I'm pretty sure it is defunct.

;; This next block all imported en-mass from the api branch
(define *http-requests-in-progress* 0)
(define *http-connections-next-cleanup* (current-seconds))

(define (http-transport:get-time-to-cleanup)
  (let ((res #f))
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246

247
248

249

250
251
252



253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374


375
376
377

378
379
380
381
382
383
384
385






386


387
388
389
390
391

392
393
394









395
396
397
398
399
400
401
402
403


404


405
406
407
408


409
410

411


412
413
414
415
416
417
418
419
420
421
422
423


424
425
426

427
428
429
430
431

432
433
434












435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464

465





466

467
468





469
470







471
472
473
474
475
476




477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494

495
496
497



498
499
500
501
502
503
504
505

506




507

508
509



510
511
512









513
514
515
516
517
518
519
520
521
522
523












524






525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
  (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*)))

;; (system "megatest -list-servers | grep alive || megatest -server - -daemonize && sleep 4")

;; <html>
;; <head></head>
;; <body>1 Hello, world! Goodbye Dolly</body></html>
;; Send msg to serverdat and receive result
(define (http-transport:client-send-receive serverdat msg #!key (numretries 30))
  (let* (;; (url        (http-transport:make-server-url serverdat))
	 (fullurl    (if (list? serverdat)
			 (caddr serverdat)
			 (begin
			   (debug:print 0 "FATAL ERROR: http-transport:client-send-receive called with no server info")
			   (exit 1)))) ;; (conc url "/ctrl")) ;; (conc url "/?dat=" msg)))
	 (res        #f))
    (handle-exceptions
     exn

     (begin
       (print "ERROR IN http-transport:client-send-receive " ((condition-property-accessor 'exn 'message) exn))

       (thread-sleep! 2)

       (if (> numretries 0)
	   (http-transport:client-send-receive serverdat msg numretries: (- numretries 1))))
     (begin



       (debug:print-info 11 "fullurl=" fullurl "\n")
       ;; set up the http-client here
       (max-retry-attempts 5)
       ;; consider all requests indempotent
       (retry-request? (lambda (request)
			 #t))   ;;  		 (thread-sleep! (/ (if (> numretries 100) 100 numretries) 10))
       ;; (set! numretries (- numretries 1))
       ;;  		 #t))
       ;; 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*)
			      (set! res (with-input-from-request 
					 fullurl 
					 (list (cons 'dat msg)) 
					 read-string))
			      (close-all-connections!) 
			      (mutex-unlock! *http-mutex*)))
	      (time-out     (lambda ()
			      (thread-sleep! 45)
			      (if (not res)
				  (begin
				    (debug:print 0 "WARNING: communication with the server timed out.")
				    (mutex-unlock! *http-mutex*)
				    (http-transport:client-send-receive serverdat msg numretries: (- numretries 1))
				    (if (< numretries 3) ;; on last try just exit
					(begin
					  (debug:print 0 "ERROR: communication with the server timed out. Giving up.")
					  (exit 1)))))))
	      (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 "got res=" res)
	 (let ((match (string-search (regexp "<body>(.*)<.body>") res)))
	   (debug:print-info 11 "match=" match)
	   (let ((final (cadr match)))
	     (debug:print-info 11 "final=" final)
	     final)))))))

;; Send "cmd" with json payload "params" to serverdat and receive result
;;
(define (http-transport:client-api-send-receive serverdat cmd params #!key (numretries 30))
  (let* ((fullurl    (if (list? serverdat)
			 (cadddr serverdat) ;; this is the uri for /api
			 (begin
			   (debug:print 0 "FATAL ERROR: http-transport:client-send-receive called with no server info")
			   (exit 1))))
	 (res        #f))
    (handle-exceptions
     exn
     (begin
       ;; TODO: Send this output to a log file so it isn't lost when running as daemon
       (print "ERROR IN http-transport:client-send-receive " ((condition-property-accessor 'exn 'message) exn))
       (thread-sleep! 2)
       (if (> numretries 0)
	   (http-transport:client-api-send-receive serverdat cmd params numretries: (- numretries 1))))
     (begin
       (debug:print-info 11 "fullurl=" fullurl "\n")
       ;; set up the http-client here
       (max-retry-attempts 5)
       ;; consider all requests indempotent
       (retry-request? (lambda (request)
			 #t))   ;;  		 (thread-sleep! (/ (if (> numretries 100) 100 numretries) 10))
       ;; (set! numretries (- numretries 1))
       ;;  		 #t))
       ;; send the data and get the response
       ;; extract the needed info from the http data and 
       ;; process and return it.

       ;; (with-input-from-request "http://localhost/echo-service"
       ;;                  '((test . "value")) read-string)

       (let* ((send-recieve (lambda ()
			;;       (let ((dat #f)
			;; 	    (cleanup (http-transport:get-time-to-cleanup)))
			;; 	(if cleanup 
			;; 	    (http-transport:inc-requests-and-prep-to-close-all-connections)
			;; 	    (http-transport:inc-requests-count))
			;; 	;; Do the actual data transfer NB// KEPP THIS IN SYNC WITH http-transport:client-send-receive
				 (mutex-lock! *http-mutex*)
				 (set! res (with-input-from-request ;; was dat
					   fullurl 
					   (list (cons 'key "thekey")
						 (cons 'cmd cmd)
						 (cons 'params params))
					   read-string))
				 ;; Shouldn't this be a call to the managed call-all-connections stuff above?
				(close-all-connections!)
				(mutex-unlock! *http-mutex*)
				))
	                          ;; (if cleanup
				  ;;   ;; mutex already set
				  ;;   (begin
				  ;;     (set! res dat)
				  ;;     (http-transport:dec-requests-count-and-close-all-connections))
				  ;;   (http-transport:dec-requests-count
				  ;;    (lambda ()
				  ;;      (set! res dat)))))))
	      (time-out     (lambda ()
			      (thread-sleep! 45)
			      (if (not res)
				  (begin
				    (debug:print 0 "WARNING: communication with the server timed out.")
				    (mutex-unlock! *http-mutex*)
				    (http-transport:client-api-send-receive serverdat cmd params numretries: (- numretries 1))
				    (if (< numretries 3) ;; on last try just exit
					(begin
					  (debug:print 0 "ERROR: communication with the server timed out. Giving up.")
					  (exit 1)))))))
	      (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 "got res=" res)
	 res)))))



(define (http-transport:client-connect iface port)
  (let* ((login-res   #f)
	 (uri-dat     (make-request method: 'POST uri: (uri-reference (conc "http://" iface ":" port "/ctrl"))))

	 (uri-api-dat (make-request method: 'POST uri: (uri-reference (conc "http://" iface ":" port "/api"))))
	 (serverdat   (list iface port uri-dat uri-api-dat)))
    (set! *runremote* serverdat) ;; may or may not be good ...
    (set! login-res (rmt:login))
    (if (and (list? login-res)
	     (car login-res))
	(begin
	  (debug:print-info 2 "Logged in and connected to " iface ":" port)






	  (set! *runremote* serverdat)


	  serverdat)
	(begin
	  (debug:print-info 0 "ERROR: Failed to login or connect to " iface ":" port)
	  (exit 1)))))
;; 	  (set! *runremote* #f)

;; 	  (set! *transport-type* 'fs)
;; 	  #f))))











;; 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
  (let* ((server-info (let loop ()


                        (let ((sdat #f))


                          (mutex-lock! *heartbeat-mutex*)
                          (set! sdat *runremote*)
                          (mutex-unlock! *heartbeat-mutex*)
                          (if sdat


			      sdat
                              (begin

                                (sleep 4)


                                (loop))))))
         (iface       (car server-info))
         (port        (cadr server-info))
         (last-access 0)
	 (tdb         (tasks:open-db))
	 (spid        ;;(open-run-close tasks:server-get-server-id tasks:open-db #f iface port #f))
	   (tasks:server-get-server-id tdb #f iface port #f))
	 (server-timeout (let ((tmo (config-lookup  *configdat* "server" "timeout")))
			   (if (and (string? tmo)
				    (string->number tmo))
			       (* 60 60 (string->number tmo))
			       ;; default to three days


			       (* 3 24 60 60)))))
    (debug:print-info 2 "server-timeout: " server-timeout ", server pid: " spid " on " iface ":" port)
    (let loop ((count 0))

      ;; Use this opportunity to sync the inmemdb to db
      (let ((start-time (current-milliseconds))
	    (sync-time  #f)
	    (rem-time   #f))
	(if *inmemdb* (db:sync-tables (db:tbls *inmemdb*) *inmemdb* *db*)) ;; (db:sync-to *inmemdb* *db*))

	(set! sync-time  (- (current-milliseconds) start-time))
	(debug:print 0 "SYNC: time= " sync-time)
	(set! rem-time (quotient (- 4000 sync-time) 1000))












	(if (and (< rem-time 4)
		 (> rem-time 0))
	    (thread-sleep! rem-time)))

      ;; (thread-sleep! 4) ;; no need to do this very often

      (if (< count 1) ;; 3x3 = 9 secs aprox
	  (loop (+ count 1)))
      
      ;; Check that iface and port have not changed (can happen if server port collides)
      (mutex-lock! *heartbeat-mutex*)
      (set! sdat *runremote*)
      (mutex-unlock! *heartbeat-mutex*)
      
      (if (or (not (equal? sdat (list iface port)))
	      (not spid))
	  (begin 
	    (debug:print-info 0 "interface changed, refreshing iface and port info")
	    (set! iface (car sdat))
	    (set! port  (cadr sdat))
	    (set! spid  (tasks:server-get-server-id tdb #f iface port #f))))
      
      ;; NOTE: Get rid of this mechanism! It really is not needed...
      ;; (open-run-close tasks:server-update-heartbeat tasks:open-db spid)
      (tasks:server-update-heartbeat tdb spid)
      
      ;; (if ;; (or (> numrunning 0) ;; stay alive for two days after last access
      (mutex-lock! *heartbeat-mutex*)
      (set! last-access *last-db-access*)
      (mutex-unlock! *heartbeat-mutex*)

      ;; (debug:print 11 "last-access=" last-access ", server-timeout=" server-timeout)





      (if (and *server-run*

	       (> (+ last-access server-timeout)
		  (current-seconds)))





	  (begin
	    (debug:print-info 0 "Server continuing, seconds since last db access: " (- (current-seconds) last-access))







	    (loop 0))
	  (begin
	    (debug:print-info 0 "Starting to shutdown the server.")
	    ;; need to delete only *my* server entry (future use)
	    (set! *time-to-exit* #t)
	    (if *inmemdb* (db:sync-tables (db:tbls *inmemdb*) *inmemdb* *db*)) ;; (db:sync-to *inmemdb* *db*))




	    (open-run-close tasks:server-deregister-self tasks:open-db (get-host-name))
	    (thread-sleep! 1)
	    (debug:print-info 0 "Max cached queries was    " *max-cache-size*)
	    (debug:print-info 0 "Number of cached writes   " *number-of-writes*)
	    (debug:print-info 0 "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 "Number non-cached queries "  *number-non-write-queries*)
	    (debug:print-info 0 "Average non-cached time   "
			      (if (eq? *number-non-write-queries* 0)
				  "n/a (no queries)"
				  (/ *total-non-write-delay* 
				     *number-non-write-queries*))
			      " ms")
	    (debug:print-info 0 "Server shutdown complete. Exiting")

	    (exit))))))

;; all routes though here end in exit ...



(define (http-transport:launch)
  (if (not *toppath*)
      (if (not (setup-for-run))
	  (begin
	    (debug:print 0 "ERROR: cannot find megatest.config, exiting")
	    (exit))))
  (debug:print-info 2 "Starting the standalone server")
  (if (args:get-arg "-daemonize")

      (daemon:ize))




  (let ((hostinfo (open-run-close tasks:get-best-server tasks:open-db)))

    (debug:print 11 "http-transport:launch hostinfo=" hostinfo)
    ;; #(1 "143.182.207.24" 5736 -1 "http" 22771 "hostname")



    (if hostinfo
	(debug:print-info 2 "NOT starting new server, one is already running on " (vector-ref hostinfo 1) ":" (vector-ref hostinfo 2))
	(if *toppath* 









	    (let* ((th2 (make-thread (lambda ()
				       (http-transport:run 
					(if (args:get-arg "-server")
					    (args:get-arg "-server")
					    "-"))) "Server run"))
		   (th3 (make-thread http-transport:keep-running "Keep running")))
;;		   (th1 (make-thread server:write-queue-handler  "write queue")))
	      (set! *cache-on* #t)
	      (set! *db*       (open-db))
	      (set! *inmemdb*  (open-in-mem-db))
	      (db:sync-tables (db:tbls *db*) *db* *inmemdb*) ;; (db:sync-to *db* *inmemdb*)



















	      (thread-start! th2)
	      (thread-start! th3)
	      ;; (thread-start! th1)
	      (set! *didsomething* #t)
	      (thread-join! th2))
	    (debug:print 0 "ERROR: Failed to setup for megatest")))
    (exit)))

;; (use trace)
;; (trace http-transport:keep-running 
;;        tasks:server-update-heartbeat
;;        tasks:server-get-server-id)
;;        tasks:get-best-server
;;        http-transport:run
;;        http-transport:launch
;;        http-transport:try-start-server
;;        http-transport:client-send-receive
;;        http-transport:make-server-url
;;        tasks:server-register
;;        tasks:server-delete
;;        start-server
;;        hostname->ip
;;        with-input-from-request
;;        tasks:server-deregister-self)

(define (http-transport:server-signal-handler signum)
  (handle-exceptions
   exn
   (debug:print " ... exiting ...")
   (let ((th1 (make-thread (lambda ()
			     (thread-sleep! 1))
			     ;; (if (not *received-response*)
			     ;;	 (receive-message* *runremote*))) ;; flush out last call if applicable
			   "eat response"))
	 (th2 (make-thread (lambda ()
			     (debug:print 0 "ERROR: 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 "       Done.")
			     (exit 4))
			   "exit on ^C timer")))







<
|
|
<
<
<
|
<
|
|

|
|



>
|
<
>
|
>
|
|
|
>
>
>
|

|


<
<
|





<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
|
<
|
<
<
<
<
<
<
<
<
|
|
|
|
|
|
|
|
|
|
<
<
<
<
<
<
<
<


<
<
<
<
<
<
<
<
|









>
>
|
|
<
>
|
|
<
<
|
|
|
|
>
>
>
>
>
>
|
>
>
|
|
|
<
<
>
|
|

>
>
>
>
>
>
>
>
>




|



|
>
>

>
>

|

|
>
>


>

>
>
|




<
<
|



|
>
>
|
<
|
>




|
>

<

>
>
>
>
>
>
>
>
>
>
>
>
|
|
|
|
<
|

|



|



|



|
<

<
<
<
|
<



>

>
>
>
>
>

>


>
>
>
>
>


>
>
>
>
>
>
>
|




|
>
>
>
>
|
|
















>



>
>
>
|
<
|
<
<
<
<

>
|
>
>
>
>
|
>
|
<
>
>
>
|
<
|
>
>
>
>
>
>
>
>
>
|
<
<
<
<
<
<
<
<
<
<
>
>
>
>
>
>
>
>
>
>
>
>

>
>
>
>
>
>
|
|
<
|
|
<
|
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<







<
<







231
232
233
234
235
236
237

238
239



240

241
242
243
244
245
246
247
248
249
250

251
252
253
254
255
256
257
258
259
260
261
262
263
264


265
266
267
268
269
270




























































271

272








273
274
275
276
277
278
279
280
281
282








283
284








285
286
287
288
289
290
291
292
293
294
295
296
297
298

299
300
301


302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317


318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361


362
363
364
365
366
367
368
369

370
371
372
373
374
375
376
377
378

379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395

396
397
398
399
400
401
402
403
404
405
406
407
408
409
410

411



412

413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476

477




478
479
480
481
482
483
484
485
486
487

488
489
490
491

492
493
494
495
496
497
498
499
500
501
502










503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523

524
525

526

















527
528
529
530
531
532
533


534
535
536
537
538
539
540
  (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))

  (let* ((fullurl    (if (vector? serverdat)
			 (http-transport:server-dat-get-api-req serverdat)
			 (begin
			   (debug:print 0 "FATAL ERROR: http-transport:client-api-send-receive called with no server info")
			   (exit 1))))
	 (res        #f))
    (handle-exceptions
     exn
     (if (> numretries 0)
	 (begin

	   (mutex-unlock! *http-mutex*)
	   (thread-sleep! 1)
	   (close-all-connections!)
	   (debug:print 0 "WARNING: Failed to communicate with server, trying again, numretries left: " numretries)
	   (http-transport:client-api-send-receive run-id serverdat cmd params numretries: (- numretries 1)))
	 (begin
	   (mutex-unlock! *http-mutex*)
	   #f))
     (begin
       (debug:print-info 11 "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 
       ;; 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 (with-input-from-request ;; was dat
					 fullurl 
					 (list (cons 'key "thekey")
					       (cons 'cmd cmd)
					       (cons 'params params))
					 read-string))
			      ;; 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)








			      #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 "got res=" res)
	 res)))))

;; careful closing of connections stored in *runremote*
;;
(define (http-transport:close-connections run-id)
  (let* ((server-dat (hash-table-ref/default *runremote* run-id #f)))

    (if (vector? server-dat)
	(let ((api-dat (http-transport:server-dat-get-api-uri server-dat)))
	  (close-connection! api-dat)


	  #t)
	#f)))


(define (make-http-transport:server-dat)(make-vector 5))
(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-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))

;;
;; 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)))
    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 server-id run-id)
  ;; if none running or if > 20 seconds since 
  ;; server last used then start shutdown
  ;; This thread waits for the server to come alive
  (let* ((server-info (let loop ((start-time (current-seconds))
				 (changed    #t)
				 (last-sdat  "not this"))
                        (let ((sdat #f))
			  (thread-sleep! 0.01)
			  (debug:print-info 0 "Waiting for server alive signal")
                          (mutex-lock! *heartbeat-mutex*)
                          (set! sdat *server-info*)
                          (mutex-unlock! *heartbeat-mutex*)
                          (if (and sdat
				   (not changed)
				   (> (- (current-seconds) start-time) 2))
			      sdat
                              (begin
				(debug:print-info 0 "Still waiting, last-sdat=" last-sdat)
                                (sleep 4)
                                (loop start-time
				      (equal? sdat last-sdat)
				      sdat))))))
         (iface       (car server-info))
         (port        (cadr server-info))
         (last-access 0)
	 (tdb         (tasks:open-db))


	 (server-timeout (let ((tmo (configf:lookup  *configdat* "server" "timeout")))
			   (if (and (string? tmo)
				    (string->number tmo))
			       (* 60 60 (string->number tmo))
			       ;; (* 3 24 60 60) ;; default to three days
			       (* 60 1)         ;; default to one minute
			       ;; (* 60 60 25)      ;; default to 25 hours
			       ))))

    (let loop ((count         0)
	       (server-state 'available))
      ;; Use this opportunity to sync the inmemdb to db
      (let ((start-time (current-milliseconds))
	    (sync-time  #f)
	    (rem-time   #f))

	(if *inmemdb* (db:sync-touched *inmemdb* *run-id* force-sync: #t))
	(set! sync-time  (- (current-milliseconds) start-time))

	(set! rem-time (quotient (- 4000 sync-time) 1000))
	(debug:print 2 "SYNC: time= " sync-time ", rem-time=" rem-time)

      ;;
      ;; set_running after our first pass through and start the db
      ;;
      (if (eq? server-state 'available)
	  (begin
	    (tasks:server-set-state! tdb server-id "dbprep")
	    (thread-sleep! 5) ;; give some margin for queries to complete before switching from file based access to server based access
	    (set! *inmemdb*  (db:setup run-id))
	    (tasks:server-set-state! tdb server-id "running")))

      (if (and (<= rem-time 4)
	       (> rem-time 0))
	  (thread-sleep! rem-time)
	  (thread-sleep! 4))) ;; fallback for if the math is changed ...

      
      (if (< count 1) ;; 3x3 = 9 secs aprox
	  (loop (+ count 1) 'running))
      
      ;; 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 (or (not (equal? sdat (list iface port)))
	      (not server-id))
	  (begin 
	    (debug:print-info 0 "interface changed, refreshing iface and port info")
	    (set! iface (car sdat))
	    (set! port  (cadr sdat))))

      



      ;; Transfer *last-db-access* to last-access to use in checking that we are still alive

      (mutex-lock! *heartbeat-mutex*)
      (set! last-access *last-db-access*)
      (mutex-unlock! *heartbeat-mutex*)

      ;; (debug:print 11 "last-access=" last-access ", server-timeout=" server-timeout)
      ;;
      ;; no_traffic, no running tests, if server 0, no running servers
      ;;
      ;; (let ((wait-on-running (configf:lookup *configdat* "server" "wait-on-running"))) ;; wait on running tasks (if not true then exit on time out)
      ;;
      (if (and *server-run*
	       ;; (or
	       (> (+ last-access server-timeout)
		  (current-seconds)))
;;		   (and (eq? run-id 0)
;;			(> (tasks:num-servers-non-zero-running tdb) 0))
;;		   (and (not (eq? run-id 0)) ;; only makes sense in non-zero run-id servers
;;			(> (db:get-count-tests-actually-running *inmemdb* run-id) 0))
;;		   ))
	  (begin
	    (debug:print-info 0 "Server continuing, seconds since last db access: " (- (current-seconds) last-access))
	    ;;
	    ;; Consider implementing some smarts here to re-insert the record or kill self is
	    ;; the db indicates so
	    ;;
	    ;; (if (tasks:server-am-i-the-server? tdb run-id)
	    ;;     (tasks:server-set-state! tdb server-id "running"))
	    ;;
	    (loop 0 server-state))
	  (begin
	    (debug:print-info 0 "Starting to shutdown the server.")
	    ;; need to delete only *my* server entry (future use)
	    (set! *time-to-exit* #t)
	    (if *inmemdb* (db:sync-touched *inmemdb* *run-id* force-sync: #t))
	    ;;
	    ;; start_shutdown
	    ;;
	    (tasks:server-set-state! tdb server-id "shutting-down")
	    (portlogger:open-run-close portlogger:set-port port "released")
	    (thread-sleep! 5)
	    (debug:print-info 0 "Max cached queries was    " *max-cache-size*)
	    (debug:print-info 0 "Number of cached writes   " *number-of-writes*)
	    (debug:print-info 0 "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 "Number non-cached queries "  *number-non-write-queries*)
	    (debug:print-info 0 "Average non-cached time   "
			      (if (eq? *number-non-write-queries* 0)
				  "n/a (no queries)"
				  (/ *total-non-write-delay* 
				     *number-non-write-queries*))
			      " ms")
	    (debug:print-info 0 "Server shutdown complete. Exiting")
	    (tasks:server-delete-record tdb server-id " http-transport:keep-running")
	    (exit))))))

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

  (set! *run-id*   run-id)




  (if (args:get-arg "-daemonize")
      (begin
	(daemon:ize)
	(if *alt-log-file* ;; we should re-connect to this port, I think daemon:ize disrupts it
	    (begin
	      (current-error-port *alt-log-file*)
	      (current-output-port *alt-log-file*)))))
  (if (server:check-if-running run-id)
      (begin
	(debug:print 0 "INFO: Server for run-id " run-id " already running")

	(exit 0)))
  (let loop ((server-id (tasks:server-lock-slot (tasks:get-db) run-id))
	     (remtries  4))
    (if (not server-id)

	(if (> remtries 0)
	    (begin
	      (thread-sleep! 2)
	      (loop (tasks:server-lock-slot (tasks:get-db) run-id)
		    (- remtries 1)))
	    (begin
	      ;; since we didn't get the server lock we are going to clean up and bail out
	      (debug:print-info 2 "INFO: server pid=" (current-process-id) ", hostname=" (get-host-name) " not starting due to other candidates ahead in start queue")
	      (tasks:server-delete-records-for-this-pid (tasks:get-db) " http-transport:launch")
	      ))
	(let* ((th2 (make-thread (lambda ()










				   (debug:print-info 0 "Server run thread started")
				   (http-transport:run 
				    (if (args:get-arg "-server")
					(args:get-arg "-server")
					"-")
				    run-id
				    server-id)) "Server run"))
	       (th3 (make-thread (lambda ()
				   (debug:print-info 0 "Server monitor thread started")
				   (http-transport:keep-running server-id run-id))
				 "Keep running")))
	  ;; Database connection


	  ;; don't start the db here

	  ;; (set! *inmemdb*  (db:setup run-id))


	  (thread-start! th2)
	  (thread-start! th3)

	  (set! *didsomething* #t)
	  (thread-join! th2)

	  (exit)))))


















(define (http-transport:server-signal-handler signum)
  (handle-exceptions
   exn
   (debug:print " ... exiting ...")
   (let ((th1 (make-thread (lambda ()
			     (thread-sleep! 1))


			   "eat response"))
	 (th2 (make-thread (lambda ()
			     (debug:print 0 "ERROR: 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 "       Done.")
			     (exit 4))
			   "exit on ^C timer")))