Megatest

Diff
Login

Differences From Artifact [cd536fb107]:

To Artifact [b3af46403d]:


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
	
	directory-utils
	;; http-client
	;; intarweb
	matchable
	md5
	message-digest

	(prefix base64 base64:)
	(prefix sqlite3 sqlite3:)
	regex
	s11n
	;; spiffy
	;; spiffy-directory-listing
	;; spiffy-request-vars
	srfi-1
	srfi-13
	srfi-18
	srfi-69
	stack
	system-information
	tcp6
	typed-records
	uri-common
	z3
       
	apimod
	commonmod
	configfmod







>













|







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
	
	directory-utils
	;; http-client
	;; intarweb
	matchable
	md5
	message-digest
	nng ;; nanomsg
	(prefix base64 base64:)
	(prefix sqlite3 sqlite3:)
	regex
	s11n
	;; spiffy
	;; spiffy-directory-listing
	;; spiffy-request-vars
	srfi-1
	srfi-13
	srfi-18
	srfi-69
	stack
	system-information
	;; tcp6
	typed-records
	uri-common
	z3
       
	apimod
	commonmod
	configfmod
117
118
119
120
121
122
123

124
125
126
127
128
129
130

;; info about me as a server
;;
(defstruct servdat
  (host #f)
  (port #f)
  (uuid #f)

  (dbfile #f)
  (api-url #f)
  (api-uri #f)
  (api-req #f)
  (status 'starting)
  (trynum 0) ;; count the number of ports we've tried
  ) 







>







118
119
120
121
122
123
124
125
126
127
128
129
130
131
132

;; info about me as a server
;;
(defstruct servdat
  (host #f)
  (port #f)
  (uuid #f)
  (rep  #f)
  (dbfile #f)
  (api-url #f)
  (api-uri #f)
  (api-req #f)
  (status 'starting)
  (trynum 0) ;; count the number of ports we've tried
  ) 
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
		#t)
	      (start-main-srv)))
	(start-main-srv))))

;; NB// remote is a rmt:remote struct
;;
(define (rmt:general-open-connection remote apath dbname #!key (num-tries 5))
  (let ((mdbname (db:run-id->dbname #f)))

    (cond
     ((not (rmt:get-conn remote apath mdbname)) ;; no channel open to main? 

      (rmt:open-main-connection remote apath)
      (thread-sleep! 2)
      (rmt:general-open-connection remote apath mdbname))
     ((not (rmt:get-conn remote apath dbname))                 ;; no channel open to dbname?     
      (let* ((res (rmt:send-receive-real remote apath mdbname 'get-server `(,apath ,dbname))))
	(case res
	  ((server-started)
	   (if (> num-tries 0)
	       (begin
		 (thread-sleep! 2)
		 (rmt:general-open-connection remote apath dbname num-tries: (- num-tries 1)))
	       (begin
		 (debug:print-error 0 *default-log-port* "Failed to start servers needed or open channel to "apath", "dbname)
		 (exit 1))))
	  (else
	   (if (list? res) ;; server has been registered and the info was returned. pass it on.






	       res
















	       (begin
		 (debug:print-info 0 *default-log-port* "Unexpected result: " res)
		 res)))))))))




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





;; Defaults to current area
;;
(define (rmt:send-receive cmd rid params #!key (attemptnum 1)(area-dat #f))
  (if (not *rmt:remote*)(set! *rmt:remote* (make-rmt:remote)))
  (let* ((apath *toppath*)
	 (conns *rmt:remote*)
	 (dbname (db:run-id->dbname rid)))





    (rmt:general-open-connection conns apath dbname)
    (rmt:send-receive-real conns apath dbname cmd params)))

#;(define (rmt:send-receive-setup conn)
  (if (not (rmt:conn-inport conn))
      (let-values (((i o) (tcp-connect (rmt:conn-ipaddr conn)
				       (rmt:conn-port conn))))
	(rmt:conn-inport-set! conn i)
	(rmt:conn-outport-set! conn o))))
  
;; db is at apath/.db/dbname, rid is an intermediary solution and will be removed
;; sometime in the future
;;
(define (rmt:send-receive-real remote apath dbname cmd params)
  (let* ((conn (rmt:get-conn remote apath dbname)))
    (assert conn "FATAL: rmt:send-receive-real called without the needed channels opened")
    (pp (rmt:conn->alist conn))
    ;; (rmt:send-receive-setup conn)
    (let-values (((i o) (tcp-connect (rmt:conn-ipaddr conn)
				     (rmt:conn-port conn))))
      (let* ((key     #f)
	     (payload `((cmd    . ,cmd)
			(key    . ,(rmt:conn-srvkey conn))
			(params . ,params)))
	     (res      (begin
			 (write payload o) ;; (rmt:conn-outport conn))
			 (with-input-from-port
			     i ;; (rmt:conn-inport conn)
			   read))))
	(close-input-port i)
	(close-output-port o)
	res))))
;;  (if (string? res)
;; 	  (string->sexpr res)
;; 	  res))))



;; db is at apath/.db/dbname, rid is an intermediary solution and will be removed
;; sometime in the future.
;;
;; Purpose - call the main.db server and request a server be started
;; for the given area path and dbname
;;
;; (define (rmt:send-receive-server-start remote apath dbname)
;;   (let* ((conn (rmt:get-conn remote apath dbname)))
;;     (assert conn "FATAL: Unable to connect to db "apath"/"dbname)
;;     #;(let* ((res      (with-input-from-request
;; 		      (rmt:conn->uri conn "api") 
;; 		      `((params . (,apath ,dbname)))
;; 		      read-string)))
;;       (string->sexpr res))))

(define (rmt:print-db-stats)
  (let ((fmtstr "~40a~7-d~9-d~20,2-f")) ;; "~20,2-f"
    (debug:print 18 *default-log-port* "DB Stats\n========")
    (debug:print 18 *default-log-port* (format #f "~40a~8a~10a~10a" "Cmd" "Count" "TotTime" "Avg"))
    (for-each (lambda (cmd)
		(let ((cmd-dat (hash-table-ref *db-stats* cmd)))
		  (debug:print 18 *default-log-port* (format #f fmtstr cmd (vector-ref cmd-dat 0) (vector-ref cmd-dat 1) (/ (vector-ref cmd-dat 1)(vector-ref cmd-dat 0))))))
	      (sort (hash-table-keys *db-stats*)
		    (lambda (a b)
		      (> (vector-ref (hash-table-ref *db-stats* a) 0)







|
>

|
>

<














>
>
>
>
>
>
|
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>


|
>
>
>



>
>
>





|
|
|
>
>
>
>
>
|
|














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


















|







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
		#t)
	      (start-main-srv)))
	(start-main-srv))))

;; NB// remote is a rmt:remote struct
;;
(define (rmt:general-open-connection remote apath dbname #!key (num-tries 5))
  (let* ((mdbname (db:run-id->dbname #f))
	 (mconn   (rmt:get-conn remote apath mdbname)))
    (cond
     ((or (not mconn) ;; no channel open to main?
	  (< (rmt:conn-expires mconn)(+ (current-seconds) 2))) ;; restablish connection if less than 2 seconds on the lease
      (rmt:open-main-connection remote apath)

      (rmt:general-open-connection remote apath mdbname))
     ((not (rmt:get-conn remote apath dbname))                 ;; no channel open to dbname?     
      (let* ((res (rmt:send-receive-real remote apath mdbname 'get-server `(,apath ,dbname))))
	(case res
	  ((server-started)
	   (if (> num-tries 0)
	       (begin
		 (thread-sleep! 2)
		 (rmt:general-open-connection remote apath dbname num-tries: (- num-tries 1)))
	       (begin
		 (debug:print-error 0 *default-log-port* "Failed to start servers needed or open channel to "apath", "dbname)
		 (exit 1))))
	  (else
	   (if (list? res) ;; server has been registered and the info was returned. pass it on.
	       (begin ;;  ("192.168.0.9" 53817
		      ;;  "5e34239f48e8973b3813221e54701a01" "24310"
		      ;;  "192.168.0.9"
		      ;;  "/home/matt/data/megatest/tests/simplerun"
		 ;;  ".db/1.db")
		 (match
		  res
		  ((host port servkey pid ipaddr apath dbname)
		   (debug:print-info 0 *default-log-port* "got "res)
		   (hash-table-set! (rmt:remote-conns remote)
				    dbname
				    (make-rmt:conn
				     apath: apath
				     dbname: dbname
				     hostport: (conc host":"port)
				     ipaddr: ipaddr
				     port: port
				     srvkey: servkey
				     lastmsg: (current-seconds)
				     expires: (+ (current-seconds) 60))))
		  (else
		   (debug:print-info 0 *default-log-port* "return data from starting server did not match host port servkey pid ipaddr apath dbname " res)))
		 res)
	       (begin
		 (debug:print-info 0 *default-log-port* "Unexpected result: " res)
		 res))))))
     

     )))

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

;; FOR DEBUGGING SET TO #t
(define *localmode* #t)
(define *dbstruct* (make-dbr:dbstruct))

;; Defaults to current area
;;
(define (rmt:send-receive cmd rid params #!key (attemptnum 1)(area-dat #f))
  (if (not *rmt:remote*)(set! *rmt:remote* (make-rmt:remote)))
  (let* ((apath      *toppath*)
	 (conns      *rmt:remote*)
	 (dbname     (db:run-id->dbname rid)))
    (if *localmode*
	(let* ((dbdat    (dbr:dbstruct-get-dbdat *dbstruct* dbname))
	       (indat    `((cmd . ,cmd)(params . ,params))))
	  (api:process-request *dbstruct* indat))
	(begin
	  (rmt:general-open-connection conns apath dbname)
	  (rmt:send-receive-real conns apath dbname cmd params)))))

#;(define (rmt:send-receive-setup conn)
  (if (not (rmt:conn-inport conn))
      (let-values (((i o) (tcp-connect (rmt:conn-ipaddr conn)
				       (rmt:conn-port conn))))
	(rmt:conn-inport-set! conn i)
	(rmt:conn-outport-set! conn o))))
  
;; db is at apath/.db/dbname, rid is an intermediary solution and will be removed
;; sometime in the future
;;
(define (rmt:send-receive-real remote apath dbname cmd params)
  (let* ((conn (rmt:get-conn remote apath dbname)))
    (assert conn "FATAL: rmt:send-receive-real called without the needed channels opened")
    (let* ((key     #f)

	   (host    (rmt:conn-ipaddr conn))
	   (port    (rmt:conn-port   conn))

	   (payload `((cmd    . ,cmd)
		      (key    . ,(rmt:conn-srvkey conn))
		      (params . ,params)))
	   (res      (open-send-receive-nn (conc host":"port)



					   (sexpr->string payload))))




      (string->sexpr res))))




;; db is at apath/.db/dbname, rid is an intermediary solution and will be removed
;; sometime in the future.
;;
;; Purpose - call the main.db server and request a server be started
;; for the given area path and dbname
;;
;; (define (rmt:send-receive-server-start remote apath dbname)
;;   (let* ((conn (rmt:get-conn remote apath dbname)))
;;     (assert conn "FATAL: Unable to connect to db "apath"/"dbname)
;;     #;(let* ((res      (with-input-from-request
;; 		      (rmt:conn->uri conn "api") 
;; 		      `((params . (,apath ,dbname)))
;; 		      read-string)))
;;       (string->sexpr res))))

(define (rmt:print-db-stats)
  (let ((fmtstr "~40a~7-d~9-d~20,2-f")) ;; "~20,2-f"
    (debug:print 18 *default-log-port* "DB Stats, "(seconds->year-week/day-time (current-seconds))"\n=====================")
    (debug:print 18 *default-log-port* (format #f "~40a~8a~10a~10a" "Cmd" "Count" "TotTime" "Avg"))
    (for-each (lambda (cmd)
		(let ((cmd-dat (hash-table-ref *db-stats* cmd)))
		  (debug:print 18 *default-log-port* (format #f fmtstr cmd (vector-ref cmd-dat 0) (vector-ref cmd-dat 1) (/ (vector-ref cmd-dat 1)(vector-ref cmd-dat 0))))))
	      (sort (hash-table-keys *db-stats*)
		    (lambda (a b)
		      (> (vector-ref (hash-table-ref *db-stats* a) 0)
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
;; Use the special run-id == #f scenario here since there is no run yet
(define (rmt:register-run keyvals runname state status user contour)
  ;; first register in main.db (thus the #f)
  (let* ((run-id (rmt:send-receive 'register-run #f (list keyvals runname state status user contour))))
    ;; now register in the run db itself

    ;; NEED A RECORD INSERT INCLUDING SETTING id
    (rmt:send-receive 'register-run run-id (list keyvals runname state status user contour))
    
    run-id))
  
(define (rmt:get-run-name-from-id run-id)
  (rmt:send-receive 'get-run-name-from-id run-id (list run-id)))

(define (rmt:delete-run run-id)
  (rmt:send-receive 'delete-run run-id (list run-id)))







|
<







699
700
701
702
703
704
705
706

707
708
709
710
711
712
713
;; Use the special run-id == #f scenario here since there is no run yet
(define (rmt:register-run keyvals runname state status user contour)
  ;; first register in main.db (thus the #f)
  (let* ((run-id (rmt:send-receive 'register-run #f (list keyvals runname state status user contour))))
    ;; now register in the run db itself

    ;; NEED A RECORD INSERT INCLUDING SETTING id
    (rmt:send-receive 'insert-run run-id (list run-id keyvals runname state status user contour))

    run-id))
  
(define (rmt:get-run-name-from-id run-id)
  (rmt:send-receive 'get-run-name-from-id run-id (list run-id)))

(define (rmt:delete-run run-id)
  (rmt:send-receive 'delete-run run-id (list run-id)))
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
  ;; (if (rmt:send-receive 'have-incompletes? run-id (list run-id ovr-deadtime))
  (rmt:send-receive 'mark-incomplete run-id (list run-id ovr-deadtime))
  ) ;; )

(define (rmt:get-main-run-stats run-id)
  (rmt:send-receive 'get-main-run-stats #f (list run-id)))

(define (rmt:get-var varname)
  (rmt:send-receive 'get-var #f (list varname)))

(define (rmt:del-var varname)
  (rmt:send-receive 'del-var #f (list varname)))

(define (rmt:set-var varname value)
  (rmt:send-receive 'set-var #f (list varname value)))

(define (rmt:inc-var varname)
  (rmt:send-receive 'inc-var #f (list varname)))

(define (rmt:dec-var varname)
  (rmt:send-receive 'dec-var #f (list varname)))

(define (rmt:add-var varname value)
  (rmt:send-receive 'add-var #f (list varname value)))

;;======================================================================
;; M U L T I R U N   Q U E R I E S
;;======================================================================

;; Need to move this to multi-run section and make associated changes
(define (rmt:find-and-mark-incomplete-all-runs #!key (ovr-deadtime #f))







|
|

|
|

|
|

|
|

|
|

|
|







759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
  ;; (if (rmt:send-receive 'have-incompletes? run-id (list run-id ovr-deadtime))
  (rmt:send-receive 'mark-incomplete run-id (list run-id ovr-deadtime))
  ) ;; )

(define (rmt:get-main-run-stats run-id)
  (rmt:send-receive 'get-main-run-stats #f (list run-id)))

(define (rmt:get-var run-id varname)
  (rmt:send-receive 'get-var run-id (list run-id varname)))

(define (rmt:del-var run-id varname)
  (rmt:send-receive 'del-var run-id (list run-id varname)))

(define (rmt:set-var run-id varname value)
  (rmt:send-receive 'set-var run-id (list run-id varname value)))

(define (rmt:inc-var run-id varname)
  (rmt:send-receive 'inc-var #f (list run-id varname)))

(define (rmt:dec-var run-id varname)
  (rmt:send-receive 'dec-var run-id (list run-id varname)))

(define (rmt:add-var run-id varname value)
  (rmt:send-receive 'add-var run-id (list run-id varname value)))

;;======================================================================
;; M U L T I R U N   Q U E R I E S
;;======================================================================

;; Need to move this to multi-run section and make associated changes
(define (rmt:find-and-mark-incomplete-all-runs #!key (ovr-deadtime #f))
814
815
816
817
818
819
820

821
822
823
824
825
826
827
828
829
;;  1. Do a remote call to get the test path
;;  2. Continue as above
;; 
;;(define (rmt:get-steps-for-test run-id test-id)
;;  (rmt:send-receive 'get-steps-data run-id (list test-id)))

(define (rmt:teststep-set-status! run-id test-id teststep-name state-in status-in comment logfile)

  (let* ((state     (items:check-valid-items "state" state-in))
	 (status    (items:check-valid-items "status" status-in)))
    (if (or (not state)(not status))
	(debug:print 3 *default-log-port* "WARNING: Invalid " (if status "status" "state")
		     " value \"" (if status state-in status-in) "\", update your validvalues section in megatest.config"))
    (rmt:send-receive 'teststep-set-status! run-id (list run-id test-id teststep-name state-in status-in comment logfile))))


(define (rmt:delete-steps-for-test! run-id test-id)







>
|
|







837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
;;  1. Do a remote call to get the test path
;;  2. Continue as above
;; 
;;(define (rmt:get-steps-for-test run-id test-id)
;;  (rmt:send-receive 'get-steps-data run-id (list test-id)))

(define (rmt:teststep-set-status! run-id test-id teststep-name state-in status-in comment logfile)
  (let* ((valid-values (configf:lookup *configdat* "validvalues" "state"))
	 (state        (items:check-valid-items valid-values "state" state-in))
	 (status       (items:check-valid-items valid-values "status" status-in)))
    (if (or (not state)(not status))
	(debug:print 3 *default-log-port* "WARNING: Invalid " (if status "status" "state")
		     " value \"" (if status state-in status-in) "\", update your validvalues section in megatest.config"))
    (rmt:send-receive 'teststep-set-status! run-id (list run-id test-id teststep-name state-in status-in comment logfile))))


(define (rmt:delete-steps-for-test! run-id test-id)
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
	  (hash-table-set! *pre-reqs-met-cache* key (vector (current-seconds) newres))
	  newres))))

;;======================================================================
;; from metadat lookup MEGATEST_VERSION
;;
(define (common:get-last-run-version) ;; RADT => How does this work in send-receive function??; assume it is the value saved in some DB
  (rmt:get-var "MEGATEST_VERSION"))

(define (common:get-last-run-version-number)
  (string->number 
   (substring (common:get-last-run-version) 0 6)))

(define (common:set-last-run-version)
  (rmt:set-var "MEGATEST_VERSION" (common:version-signature)))

;;======================================================================
;; faux-lock is deprecated. Please use simple-lock below
;;
(define (common:faux-lock keyname #!key (wait-time 8)(allow-lock-steal #t))
  (if (rmt:no-sync-get/default keyname #f) ;; do not be tempted to compare to pid. locking is a one-shot action, if already locked for this pid it doesn't actually count
      (if (> wait-time 0)







|






|







1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
	  (hash-table-set! *pre-reqs-met-cache* key (vector (current-seconds) newres))
	  newres))))

;;======================================================================
;; from metadat lookup MEGATEST_VERSION
;;
(define (common:get-last-run-version) ;; RADT => How does this work in send-receive function??; assume it is the value saved in some DB
  (rmt:get-var #f "MEGATEST_VERSION"))

(define (common:get-last-run-version-number)
  (string->number 
   (substring (common:get-last-run-version) 0 6)))

(define (common:set-last-run-version)
  (rmt:set-var #f "MEGATEST_VERSION" (common:version-signature)))

;;======================================================================
;; faux-lock is deprecated. Please use simple-lock below
;;
(define (common:faux-lock keyname #!key (wait-time 8)(allow-lock-steal #t))
  (if (rmt:no-sync-get/default keyname #f) ;; do not be tempted to compare to pid. locking is a one-shot action, if already locked for this pid it doesn't actually count
      (if (> wait-time 0)
1441
1442
1443
1444
1445
1446
1447















































1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460

1461

1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475

1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492

1493
1494
1495
1496
1497
1498
1499
1500
1501

1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
(define (common:version-changed?)
  (not (equal? (common:get-last-run-version)
               (common:version-signature))))

(define (common:api-changed?)
  (not (equal? (substring (->string megatest-version) 0 4)
               (substring (conc (common:get-last-run-version)) 0 4))))
















































(define (std-exit-procedure)
  ;;(common:telemetry-log-close)
  (on-exit (lambda () 0))
  ;;(debug:print-info 13 *default-log-port* "std-exit-procedure called; *time-to-exit*="*time-to-exit*)
  (let ((no-hurry  (if (bdat-time-to-exit *bdat*) ;; hurry up
		       #f
		       (begin
			 (bdat-time-to-exit-set! *bdat* #t)
			 #t))))
    (debug:print-info 4 *default-log-port* "starting exit process, finalizing databases.")
    (if (and no-hurry (debug:debug-mode 18))
	(rmt:print-db-stats))

    (let ((th1 (make-thread (lambda () ;; thread for cleaning up, give it five seconds

                              (if *server-info*
				  (let ((pkt-file (conc (get-pkts-dir *toppath*)
							"/" (servdat-uuid *server-info*)
							".pkt"))
					(dbfile   (servdat-dbfile *server-info*)))
				    (if dbfile
					(begin

					  ;; do a final sync here
					  
					(if (string-match ".*/main.db$" dbfile)
					    (begin
					      (debug:print-info 0 *default-log-port* "removing pkt "pkt-file)
					      (delete-file* pkt-file)

					      (debug:print-info 0 *default-log-port* "Releasing lock for "dbfile)
					      (db:with-lock-db (servdat-dbfile *server-info*)
							       (lambda (dbh dbfile)
								 (db:release-lock dbh dbfile))))
					    (let* ((sdat *server-info*)) ;; we have a run-id server
					      (rmt:send-receive-real *rmt:remote* *toppath*
								     (db:run-id->dbname #f)
								     'deregister-server
								     `(,(servdat-uuid sdat)
								       ,(current-process-id)
								       ,(servdat-host sdat)   ;; iface
								       ,(servdat-port sdat)))))))))
			      (if *dbstruct-db* (db:close-all *dbstruct-db*)) ;; one second allocated
			      (if (bdat-task-db *bdat*)    ;; TODO: Check that this is correct for task db
				  (let ((db (cdr (bdat-task-db *bdat*))))
				    (if (sqlite3:database? db)
					(begin

					  (sqlite3:interrupt! db)
					  (sqlite3:finalize! db #t)
					  (bdat-task-db-set! *bdat* #f)))))
                              #;(http-client#close-idle-connections!)
                              (if (not (eq? *default-log-port* (current-error-port)))
                                  (close-output-port *default-log-port*))
			      (set! *default-log-port* (current-error-port))) "Cleanup db exit thread"))
	  (th2 (make-thread (lambda ()
			      (debug:print 4 *default-log-port* "Attempting clean exit. Please be patient and wait a few seconds...")

			      (if no-hurry
                                  (begin
                                    (thread-sleep! 5)) ;; give the clean up few seconds to do it's stuff
                                  (begin
      				  (thread-sleep! 2)))
      			      (debug:print 4 *default-log-port* " ... done")
      			      )
			    "clean exit")))
      (thread-start! th1)
      (thread-start! th2)
      (thread-join! th1)
      )







>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>













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

|
>




|







1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535






1536



1537
1538

1539
1540











1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
(define (common:version-changed?)
  (not (equal? (common:get-last-run-version)
               (common:version-signature))))

(define (common:api-changed?)
  (not (equal? (substring (->string megatest-version) 0 4)
               (substring (conc (common:get-last-run-version)) 0 4))))

(define (rmt:server-shutdown)
  (let ((dbfile   (servdat-dbfile *server-info*)))
    (debug:print-info 0 *default-log-port* "dbfile is "dbfile)
    (if dbfile
	(let* ((am-server  (args:get-arg "-server"))
	       (dbfile     (args:get-arg "-db"))
	       (apath      *toppath*)
	       (dbdat      (db:get-dbdat *dbstruct-db* apath dbfile))
	       (db         (dbr:dbdat-db dbdat))
	       (inmem      (dbr:dbdat-db dbdat))
	       )
	  ;; do a final sync here
	  (debug:print-info 0 *default-log-port* "Doing final sync for "apath" "dbfile" at "(current-seconds))
	  (db:sync-inmem->disk *dbstruct-db* apath dbfile force-sync: #t)
	  ;; let's finalize here
	  (debug:print-info 0 *default-log-port* "Finalizing db and inmem")
	  (sqlite3:finalize! db)
	  (sqlite3:finalize! inmem)
	  (debug:print-info 0 *default-log-port* "Finalizing db and inmem complete")
	  (if am-server
	      (if (string-match ".*/main.db$" dbfile)
		  (let ((pkt-file (conc (get-pkts-dir *toppath*)
					"/" (servdat-uuid *server-info*)
					".pkt")))
		    (debug:print-info 0 *default-log-port* "removing pkt "pkt-file)
		    (delete-file* pkt-file)
		    (debug:print-info 0 *default-log-port* "Releasing lock for "dbfile)
		    (db:with-lock-db (servdat-dbfile *server-info*)
				     (lambda (dbh dbfile)
				       (db:release-lock dbh dbfile))))
		  (let* ((sdat *server-info*) ;; we have a run-id server
			 (host (servdat-host sdat))
			 (port (servdat-port sdat))
			 (uuid (servdat-uuid sdat)))
		    (if (not (string-match ".db/main.db" (args:get-arg "-db")))
			(let* ((res (rmt:deregister-server *rmt:remote* ;; TODO/BUG: why is this requiring *rmt:remote*?
							   *toppath*
							   (servdat-host *server-info*)   ;; iface
							   (servdat-port *server-info*)
							   (servdat-uuid *server-info*)
							   (current-process-id)
							   )))
			  (debug:print-info 0 *default-log-port* "deregistered-server, res="res)))
		    
		    (debug:print-info 0 *default-log-port* "deregistering server "host":"port" with uuid "uuid)
		    )))))))

(define (std-exit-procedure)
  ;;(common:telemetry-log-close)
  (on-exit (lambda () 0))
  ;;(debug:print-info 13 *default-log-port* "std-exit-procedure called; *time-to-exit*="*time-to-exit*)
  (let ((no-hurry  (if (bdat-time-to-exit *bdat*) ;; hurry up
		       #f
		       (begin
			 (bdat-time-to-exit-set! *bdat* #t)
			 #t))))
    (debug:print-info 4 *default-log-port* "starting exit process, finalizing databases.")
    (if (and no-hurry (debug:debug-mode 18))
	(rmt:print-db-stats))
    (let ((th1 (make-thread
		(lambda () ;; thread for cleaning up, give it five seconds
		  (let* ((start-time (current-seconds)))
		    (if (and *server-info*






			     *unclean-shutdown*)



			(begin
			  (debug:print-info 0 *default-log-port* "Unclean server exit, calling server-shtudown")

			  (rmt:server-shutdown)))
		    (debug:print-info 0 *default-log-port* "Shutdown activities completed in "(- (current-seconds) start-time)" seconds"))











		  ;; (if *dbstruct-db* (db:close-all *dbstruct-db*)) ;; one second allocated
		  #;(if (bdat-task-db *bdat*)    ;; TODO: Check that this is correct for task db
		  (let ((db (cdr (bdat-task-db *bdat*))))
		  (if (sqlite3:database? db)
		  (begin
		  (debug:print-info 0 *default-log-port* "Closing down task db "db)
		  (sqlite3:interrupt! db)
		  (sqlite3:finalize! db #t)
		  (bdat-task-db-set! *bdat* #f)))))
		  #;(http-client#close-idle-connections!)
		  (if (not (eq? *default-log-port* (current-error-port)))
		      (close-output-port *default-log-port*))
		  (set! *default-log-port* (current-error-port))) "Cleanup db exit thread"))
	  (th2 (make-thread (lambda ()
			      (debug:print 4 *default-log-port* "Attempting clean exit. Mode="(if no-hurry "no-hurry" "normal")
					   " Please be patient and wait a few seconds...")
			      (if no-hurry
                                  (begin
                                    (thread-sleep! 5)) ;; give the clean up few seconds to do it's stuff
                                  (begin
				    (thread-sleep! 2)))
      			      (debug:print 4 *default-log-port* " ... done")
      			      )
			    "clean exit")))
      (thread-start! th1)
      (thread-start! th2)
      (thread-join! th1)
      )
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598

1599

1600
1601
1602
1603
1604
1605
1606
1607
1608
1609


1610
1611


















1612



1613
1614
1615
1616
1617
1618
1619
1620

1621
1622
1623
1624
1625





















1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
      #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 (rmt:launch-server hostn port)
   (if *server-info*
	(begin
	  (servdat-host-set! *server-info* hostn)
	  (servdat-port-set! *server-info* port)
	  (servdat-status-set! *server-info* 'trying-port)
	  (servdat-trynum-set! *server-info* (+ (servdat-trynum *server-info*) 1)))
	(set! *server-info* (make-servdat host: ipaddrstr port: portnum)))
   (let* ((l        (tcp-listen port))
	 (dbstruct #f))
    (let-values (((i o) (tcp-accept l)))
      ;; (write-line "Hello!" o)
      (let loop ((indat (read i)))
	(let* ((res (api:process-request dbstruct indat)))
	  (case res
	    ((quit)
	     (close-input-port i)
	     (close-output-port o))
	    (else
	     (write res o))))))))

(define (rmt: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))) 
	 (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: " port)
    (if *server-info*
	(begin
	  (servdat-host-set! *server-info* ipaddrstr)
	  (servdat-port-set! *server-info* port)
	  (servdat-status-set! *server-info* 'trying-port)
	  (servdat-trynum-set! *server-info* (+ (servdat-trynum *server-info*) 1)))
	(set! *server-info* (make-servdat host: ipaddrstr port: port)))
    (let* ((l        (rmt:try-start-server ipaddrstr port)))
      (let oloop ()
	(let-values (((i o) (tcp-accept l)))

	  ;; (write-line "Hello!" o)

	  (let loop ((indat (read i)))
	    (if (eof-object? indat)
		(begin
		  (close-input-port i)
		  (close-output-port o)
		  (oloop))
		(let* ((res (api:process-request *dbstruct-db* indat)))
		  (set! *db-last-access* (current-seconds))
		  (write res o)
		  (loop (read i))))))))


    (let* ((portnum (servdat-port *server-info*)))
      (portlogger:open-run-close portlogger:set-port portnum "released")


















      (debug:print 1 *default-log-port* "INFO: server has been stopped"))))




(define (rmt:try-start-server ipaddrstr portnum)
  (if *server-info*
      (begin
	(servdat-host-set! *server-info* ipaddrstr)
	(servdat-port-set! *server-info* portnum)
	(servdat-status-set! *server-info* 'trying-port)
	(servdat-trynum-set! *server-info* (+ (servdat-trynum *server-info*) 1)))

      (set! *server-info* (make-servdat host: ipaddrstr port: portnum)))
  (debug:print-info 0 *default-log-port* "rmt:try-start-server time="
		    (seconds->time-string (current-seconds))
		    " ipaddrsstr=" ipaddrstr
		    " portnum=" portnum)





















  (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)
	   (rmt: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
   (if *server-info*
       (servdat-status-set! *server-info* 'starting)
       (set! *server-info* (make-servdat host: ipaddrstr port: portnum)))
   
   (debug:print 0 *default-log-port* "INFO: Trying to start server on " ipaddrstr ":" portnum)
   (tcp-listen portnum)))

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

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







<
<
<
<



<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<














|









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


>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
|
>
>
>


|




|
>





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







1592
1593
1594
1595
1596
1597
1598




1599
1600
1601





















1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634




1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714




1715

1716
1717
1718
1719
1720
1721
1722
1723
      #f
      (conc "http://" (car hostport) ":" (cadr hostport))))

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





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






















(define (rmt: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))) 
	 (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: " port)
    (if *server-info*
	(begin
	  (servdat-host-set! *server-info* ipaddrstr)
	  (servdat-port-set! *server-info* port)
	  (servdat-status-set! *server-info* 'trying-port)
	  (servdat-trynum-set! *server-info* (+ (servdat-trynum *server-info*) 1)))
	(set! *server-info* (make-servdat host: ipaddrstr port: port)))
    (let* ((rep (rmt:try-start-server ipaddrstr port)))
      (let loop ((instr (nng-recv rep)))
	(let* ((data   (string->sexpr instr))
	       (res    (case data
			 ((quit) 'quit)
			 (else    (api:process-request *dbstruct-db* data))))
	       (resdat (sexpr->string res)))
	  (if (not (eq? res 'quit))
	      (begin




		(set! *db-last-access* (current-seconds))
		(nng-send rep resdat)
		(loop (nng-recv rep)))))))
    (debug:print-info 0 *default-log-port* "After server, should never see this")
    ;; server exit stuff here
    (let* ((portnum (servdat-port *server-info*)))
      (portlogger:open-run-close portlogger:set-port portnum "released")
      (rmt:server-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") ;; done in rmt:run
      ;; (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)
      (debug:print-info 0 *default-log-port* "Server shutdown complete. Exiting")
      )))

(define (rmt:try-start-server ipaddrstr portnum)
  (if *server-info* ;; update the server info as we might be trying next port
      (begin
	(servdat-host-set! *server-info* ipaddrstr)
	(servdat-port-set! *server-info* portnum)
	(servdat-status-set! *server-info* 'trying-port)
	(servdat-trynum-set! *server-info*
			     (+ (servdat-trynum *server-info*) 1)))
      (set! *server-info* (make-servdat host: ipaddrstr port: portnum)))
  (debug:print-info 0 *default-log-port* "rmt:try-start-server time="
		    (seconds->time-string (current-seconds))
		    " ipaddrsstr=" ipaddrstr
		    " portnum=" portnum)
  (if (is-port-in-use portnum)
      (begin
	(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)
	(rmt:try-start-server ipaddrstr
			      (portlogger:open-run-close
			       portlogger:find-port)))
      (begin
	(if (not *server-info*)
	    (set! *server-info* (make-servdat
				 host: ipaddrstr
				 port: portnum)))
	(servdat-status-set! *server-info* 'starting)
	(servdat-port-set!   *server-info* portnum)
	(if (not (servdat-rep *server-info*))
	    (let ((rep  (make-rep-socket)))
	      (servdat-rep-set!    *server-info* rep)
	      (socket-set! rep 'nng/recvtimeo 2000)))
	(let* ((rep (servdat-rep *server-info*)))
	  (debug:print 0 *default-log-port* "INFO: Trying to start server on " ipaddrstr ":" portnum)
	  (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)
		   (rmt:try-start-server ipaddrstr
					 (portlogger:open-run-close portlogger:find-port)))
		 (begin
		   (print "ERROR: Tried and tried but could not start the server, stopping at port "portnum))))




	   (nng-listen rep (conc "tcp://*:" portnum))

	   rep)))))

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

;;======================================================================
;; C L I E N T S
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809



1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
	 all-pkt-files)))

(define (server-address srv-pkt)
  (conc (alist-ref 'host srv-pkt) ":"
	(alist-ref 'port srv-pkt)))
	
(define (server-ready? host port key) ;; server-address is host:port
  (let-values (((i o)(handle-exceptions
		      exn
		      (values #f #f)
		      (tcp-connect host port))))
    (if (and i o)
	(begin
	  (write `((cmd . ping)
		   (key . ,key)
		   (params . ())) o)



	  (let ((res (with-input-from-port i
		       read)))
	    (close-output-port o)
	    (close-input-port i)
	    res))
;; 	    (if (string? res)
;; 		(string->sexpr res)
;; 		res)))
	(begin ;; connection failed
	  (debug:print-info 0 *default-log-port* "Server at "host":"port" is not responding.")
	  #f))))
 	      
;; (define (loop-test host port data) ;; server-address is host:port
;;   ;; ping the server and ask it
;;   ;; if it ready
;;   ;; (let* ((sdat (servdat-init #f host port #f)))
;;   ;;   (http-transport:send-receive sdat "abc" 'ping '())))
;;   (let* ((payload (sexpr->string data))
;; 	 (res     (with-input-from-request







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



|
|
|
|







1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873

1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
	 all-pkt-files)))

(define (server-address srv-pkt)
  (conc (alist-ref 'host srv-pkt) ":"
	(alist-ref 'port srv-pkt)))
	
(define (server-ready? host port key) ;; server-address is host:port
;;  (let-values (((i o)(handle-exceptions
;;		      exn
;;		      (values #f #f)
;;		      (tcp-connect host port))))
;;    (if (and i o)

  (let* ((data (sexpr->string  `((cmd . ping)
				 (key . ,key)
				 (params . ()))))
	 (res  (open-send-receive-nn (conc host ":" port) data)))
    (string->sexpr res)))

;; (let ((res (with-input-from-port i
;; 		       read)))
;; 	    (close-output-port o)
;; 	    (close-input-port i)
;; 	    res))
;; 	    (if (string? res)
;; 		(string->sexpr res)
;; 		res)))
;;     (begin ;; connection failed
;; 	  (debug:print-info 0 *default-log-port* "Server at "host":"port" is not responding.")
;; 	  #f))))

;; (define (loop-test host port data) ;; server-address is host:port
;;   ;; ping the server and ask it
;;   ;; if it ready
;;   ;; (let* ((sdat (servdat-init #f host port #f)))
;;   ;;   (http-transport:send-receive sdat "abc" 'ping '())))
;;   (let* ((payload (sexpr->string data))
;; 	 (res     (with-input-from-request
1968
1969
1970
1971
1972
1973
1974
1975
1976

1977
1978
1979
1980
1981
1982



















1983
1984
1985
1986
1987
1988
1989
		    (exit))
		  (loop start-time
			(equal? sdat last-sdat)
			sdat))))))))

(define (rmt:register-server remote apath iface port server-key dbname)
  (rmt:open-main-connection remote apath) ;; we need a channel to main.db
  (rmt:send-receive-real remote apath ;; params: host port servkey pid ipaddr dbpath
			 (db:run-id->dbname #f) 'register-server `(,iface

								   ,port
								   ,server-key
								   ,(current-process-id)
								   ,iface
								   ,apath
								   ,dbname)))




















(define (http-transport:wait-for-stable-interface #!optional (num-tries-allowed 100))
  ;; wait until *server-info* stops changing
  (let* ((stime (current-seconds)))
    (let loop ((last-host  #f)
	       (last-port  #f)
	       (tries 0))







|
|
>
|
|
|
|
|
|
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>







2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
		    (exit))
		  (loop start-time
			(equal? sdat last-sdat)
			sdat))))))))

(define (rmt:register-server remote apath iface port server-key dbname)
  (rmt:open-main-connection remote apath) ;; we need a channel to main.db
  (rmt:send-receive-real remote apath      ;; params: host port servkey pid ipaddr dbpath
			 (db:run-id->dbname #f)
			 'register-server `(,iface
					    ,port
					    ,server-key
					    ,(current-process-id)
					    ,iface
					    ,apath
					    ,dbname)))

(define (rmt:get-count-servers remote apath)
  (rmt:open-main-connection remote apath) ;; we need a channel to main.db
  (rmt:send-receive-real remote apath      ;; params: host port servkey pid ipaddr dbpath
			 (db:run-id->dbname #f)
			 'get-count-servers `(,apath
					      )))

(define (rmt:deregister-server remote apath iface port server-key dbname)
  (rmt:open-main-connection remote apath) ;; we need a channel to main.db
  (rmt:send-receive-real remote apath      ;; params: host port servkey pid ipaddr dbpath
                         (db:run-id->dbname #f)
                         'deregister-server `(,iface
                                              ,port
                                              ,server-key
                                              ,(current-process-id)
                                              ,iface
                                              ,apath
                                              ,dbname)))

(define (http-transport:wait-for-stable-interface #!optional (num-tries-allowed 100))
  ;; wait until *server-info* stops changing
  (let* ((stime (current-seconds)))
    (let loop ((last-host  #f)
	       (last-port  #f)
	       (tries 0))
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049

2050
2051
2052
2053
2054
2055
2056
    ;; main and run db servers have both got wait logic (could/should merge it)
    (if is-main
	(http-transport:wait-for-server pkts-dir dbname server-key)
	(http-transport:wait-for-stable-interface))
    ;; this is our forever loop
    (let* ((iface             (servdat-host *server-info*))
	   (port              (servdat-port *server-info*)))
      (let loop ((count         0)
		 (bad-sync-count 0)
		 (start-time     (current-milliseconds)))
	
	(if (not is-main)

	    (debug:print-info 0 *default-log-port* "servdat-status is " (servdat-status *server-info*)))

	;; set up the database handle
	(mutex-lock! *heartbeat-mutex*)
	(if (not *dbstruct-db*) ;; no db opened yet, open the db and register with main if appropriate
	    (let ((watchdog (bdat-watchdog *bdat*)))		 
	      (debug:print 0 *default-log-port* "SERVER: dbprep")







|



|
>







2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
    ;; main and run db servers have both got wait logic (could/should merge it)
    (if is-main
	(http-transport:wait-for-server pkts-dir dbname server-key)
	(http-transport:wait-for-stable-interface))
    ;; this is our forever loop
    (let* ((iface             (servdat-host *server-info*))
	   (port              (servdat-port *server-info*)))
      (let loop ((count          0)
		 (bad-sync-count 0)
		 (start-time     (current-milliseconds)))
	
	(if (and (not is-main)
		 (common:low-noise-print 60 "servdat-status"))
	    (debug:print-info 0 *default-log-port* "servdat-status is " (servdat-status *server-info*)))

	;; set up the database handle
	(mutex-lock! *heartbeat-mutex*)
	(if (not *dbstruct-db*) ;; no db opened yet, open the db and register with main if appropriate
	    (let ((watchdog (bdat-watchdog *bdat*)))		 
	      (debug:print 0 *default-log-port* "SERVER: dbprep")
2066
2067
2068
2069
2070
2071
2072



2073
2074
2075
2076
2077
2078
2079
2080
2081
2082





2083
2084
2085
2086
2087
2088
2089
			(begin 
			  (debug:print 0 *default-log-port* "We are not the server for "dbname", exiting.")
			  (exit)))))
	      (debug:print 0 *default-log-port*
			   "SERVER: running, db "dbname" opened, megatest version: "
			   (common:get-full-version))
	      ;; start the watchdog



	      (if watchdog
		  (if (not (member (thread-state watchdog)
				   '(ready running blocked
					   sleeping dead)))
		      (begin
			(debug:print-info 0 *default-log-port* "Starting watchdog thread (in state "(thread-state watchdog)")")
			(thread-start! watchdog))
		      (debug:print-info 0 *default-log-port* "Not starting watchdog thread (in state "(thread-state watchdog)")"))
		  (debug:print 0 *default-log-port* "ERROR: *watchdog* not setup, cannot start it."))
	      #;(loop (+ count 1) bad-sync-count start-time)))





	(mutex-unlock! *heartbeat-mutex*)
	
	;; 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)))







>
>
>
|








|
>
>
>
>
>







2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
			(begin 
			  (debug:print 0 *default-log-port* "We are not the server for "dbname", exiting.")
			  (exit)))))
	      (debug:print 0 *default-log-port*
			   "SERVER: running, db "dbname" opened, megatest version: "
			   (common:get-full-version))
	      ;; start the watchdog

	      ;; is this really needed?
	      
	      #;(if watchdog
		  (if (not (member (thread-state watchdog)
				   '(ready running blocked
					   sleeping dead)))
		      (begin
			(debug:print-info 0 *default-log-port* "Starting watchdog thread (in state "(thread-state watchdog)")")
			(thread-start! watchdog))
		      (debug:print-info 0 *default-log-port* "Not starting watchdog thread (in state "(thread-state watchdog)")"))
		  (debug:print 0 *default-log-port* "ERROR: *watchdog* not setup, cannot start it."))
	      #;(loop (+ count 1) bad-sync-count start-time)
	      ))

	(debug:print-info 0 *default-log-port* "syncing "*toppath*" "dbname" at "(current-seconds))
	(db:sync-inmem->disk *dbstruct-db* *toppath* dbname force-sync: #t)
	
	(mutex-unlock! *heartbeat-mutex*)
	
	;; 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)))
2101
2102
2103
2104
2105
2106
2107
2108



2109
2110
2111
2112

2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161

2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176

2177
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
2189
2190
2191
2192
2193
2194
2195
2196






































































































2197
2198
2199
2200
2201
2202
2203
	    (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)))
	    (loop 0 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
    ;;

    ;; deregister the server

    
    (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 ...
;;
;; This is the point at which servers are started
;;
(define (rmt:server-launch dbname)

  (let* ((th2 (make-thread (lambda ()
			     (debug:print-info 0 *default-log-port* "Server run thread started")
			     (rmt: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")
			     (rmt:keep-running dbname)
			     "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))

  #f
  )
	    
;; Generate a unique signature for this process, used at both client and
;; server side
(define (rmt:mk-signature)
  (message-digest-string (md5-primitive) 
			 (with-output-to-string
			   (lambda ()
			     (write (list (current-directory)
                                          (current-process-id)
					  (argv)))))))

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







































































































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

;; run ping in separate process, safest way in some cases
;;







|
>
>
>




>

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








>















>
|
|


















>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>







2200
2201
2202
2203
2204
2205
2206
2207
2208
2209
2210
2211
2212
2213
2214
2215
2216





2217

2218






2219

2220


2221












2222


2223



2224
2225
2226
2227
2228
2229
2230
2231
2232
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
2244
2245
2246
2247
2248
2249
2250
2251
2252
2253
2254
2255
2256
2257
2258
2259
2260
2261
2262
2263
2264
2265
2266
2267
2268
2269
2270
2271
2272
2273
2274
2275
2276
2277
2278
2279
2280
2281
2282
2283
2284
2285
2286
2287
2288
2289
2290
2291
2292
2293
2294
2295
2296
2297
2298
2299
2300
2301
2302
2303
2304
2305
2306
2307
2308
2309
2310
2311
2312
2313
2314
2315
2316
2317
2318
2319
2320
2321
2322
2323
2324
2325
2326
2327
2328
2329
2330
2331
2332
2333
2334
2335
2336
2337
2338
2339
2340
2341
2342
2343
2344
2345
2346
2347
2348
2349
2350
2351
2352
2353
2354
2355
2356
2357
2358
2359
2360
2361
2362
2363
2364
2365
2366
2367
2368
2369
2370
2371
2372
2373
2374
2375
2376
2377
2378
	    (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 is-main
		     (> (rmt:get-count-servers *rmt:remote* *toppath*) 1)
		     #t))
	    (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)))
	    (loop 0 bad-sync-count (current-milliseconds)))
	   (else
	    (set! *unclean-shutdown* #f)
	    (debug:print-info 0 *default-log-port* "Server timed out. seconds since last db access: " (- (current-seconds) last-access))





	    (debug:print-info 0 *default-log-port* "Starting to shutdown the server. pid="(current-process-id))

	    (rmt:server-shutdown)






	    (portlogger:open-run-close portlogger:set-port port "released")

	    (exit)


	    #;(debug:print-info 0 *default-log-port* "Sending 'quit to server, received: "












			      (open-send-receive-nn (conc iface":"port)      ;; do this here and not in server-shutdown


						    (sexpr->string 'quit)))



	    )))))))

;; Call this to start the actual server
;;
;; all routes though here end in exit ...
;;
;; This is the point at which servers are started
;;
(define (rmt:server-launch dbname)
  (debug:print-info 0 *default-log-port* "Entered rmt:server-launch")
  (let* ((th2 (make-thread (lambda ()
			     (debug:print-info 0 *default-log-port* "Server run thread started")
			     (rmt: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")
			     (rmt:keep-running dbname)
			     "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)
    (thread-join! th3)
    ;; (exit))
  )
  #f
  )
	    
;; Generate a unique signature for this process, used at both client and
;; server side
(define (rmt:mk-signature)
  (message-digest-string (md5-primitive) 
			 (with-output-to-string
			   (lambda ()
			     (write (list (current-directory)
                                          (current-process-id)
					  (argv)))))))

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

;;======================================================================
;; Nanomsg transport
;;======================================================================

(define (is-port-in-use port-num)
  (let* ((ret #f))
    (let-values (((inp oup pid)
		  (process "netstat" (list  "-tulpn" ))))
      (let loop ((inl (read-line inp)))
        (if (not (eof-object? inl))
            (begin 
	      (if (string-search (regexp (conc ":" port-num)) inl)
		  (begin
					;(print "Output: "  inl)
		    (set! ret  #t))
		  (loop (read-line inp)))))))
    ret))

;;start a server, returns the connection
;;
(define (start-nn-server portnum )
  (let ((rep (make-rep-socket))) ;; (nn-socket 'rep)))
    (socket-set! rep 'nng/recvtimeo 2000)
    (handle-exceptions
     exn
     (let ((emsg ((condition-property-accessor 'exn 'message) exn)))
       (print "ERROR: Failed to start server \"" emsg "\"")
       (exit 1))
      
     (nng-dial #;nn-bind rep (conc "tcp://*:" portnum)))
    rep))

;; open connection to server, send message, close connection
;;
(define (open-send-close-nn host-port msg #!key (timeout 3) ) ;; default timeout is 3 seconds
  (let ((req  (make-req-socket 'req))
        (uri  (conc "tcp://" host-port))
        (res  #f)
        ;; (contacts (alist-ref 'contact attrib))
        ;; (mode (alist-ref 'mode attrib))
	)
    (socket-set! req 'nng/recvtimeo 2000)
    (handle-exceptions
     exn
     (let ((emsg ((condition-property-accessor 'exn 'message) exn)))
       ;; Send notification       
       (debug:print 0 *default-log-port* "ERROR: Failed to connect / send to " uri " message was \"" emsg "\"" )
       #f)
     (nng-dial req uri)
     ;; (print "Connected to the server " )
     (nng-send req msg)
     ;; (print "Request Sent")  
     (let* ((th1  (make-thread (lambda ()
                                 (let ((resp (nng-recv req)))
                                   (nng-close! req)
                                   (set! res (if (equal? resp "ok")
                                                 #t
                                                 #f))))
                               "recv thread"))
            (th2 (make-thread (lambda ()
                                (thread-sleep! timeout)
                                (thread-terminate! th1))
			      "timer thread")))
       (thread-start! th1)
       (thread-start! th2)
       (thread-join! th1)
       res))))

(define (open-send-receive-nn host-port msg #!key (timeout 3) ) ;; default timeout is 3 seconds
  (let ((req  (make-req-socket))
        (uri  (conc "tcp://" host-port))
        (res  #f)
	;;        (contacts (alist-ref 'contact attrib))
        ;; (mode (alist-ref 'mode attrib))
	) 
    (handle-exceptions
     exn
     (let ((emsg ((condition-property-accessor 'exn 'message) exn)))
       ;; Send notification      
       (debug:print 0 *default-log-port* "ERROR: Failed to connect / send to " uri " message was \"" emsg "\", exn=" exn)
       #f)
     (nng-dial req uri)
     ;; (print "Connected to the server " )
     (nng-send req msg)
     ;; (print "Request Sent")  
     ;; receive code here
     ;;(print (nn-recv req))
     (let* ((th1  (make-thread (lambda ()
                                 (let ((resp (nng-recv req)))
                                   (nng-close! req)
                                   (print resp)
                                   (set! res resp)))
                               "recv thread"))
            (th2 (make-thread (lambda ()
                                (thread-sleep! timeout)
                                (thread-terminate! th1))
                             "timer thread")))
       (thread-start! th1)
       (thread-start! th2)
       (thread-join! th1)
       res))))

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

;; run ping in separate process, safest way in some cases
;;