| ︙ | | | ︙ | |
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
|
;; NOTE: This can fail when there is no mention of the host in /etc/hosts. FIXME
(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
(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)
(handle-exception (lambda (exn chain)
|
|
>
|
|
>
>
|
|
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
|
;; NOTE: This can fail when there is no mention of the host in /etc/hosts. FIXME
(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 area-dat)
(debug:print 2 "Attempting to start the server ...")
(let* ((configdat (megatest:area-configdat area-dat))
(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 (lambda (db)
(portlogger:find-port db area-dat))
area-dat))
(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)
(handle-exception (lambda (exn chain)
|
| ︙ | | | ︙ | |
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
|
;; 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*))
((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"))
(tdbdat (tasks:open-db)))
(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 (db:delay-if-busy tdbdat) 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
(db:delay-if-busy tdbdat)
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 (db:delay-if-busy tdbdat) 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
;;======================================================================
;;======================================================================
|
|
|
|
|
|
|
>
>
|
>
>
>
|
>
|
|
|
|
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
|
;; 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* area-dat $) ;; 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*))
((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 area-dat)))
;; This is recursively run by http-transport:run until sucessful
;;
(define (http-transport:try-start-server run-id ipaddrstr portnum server-id area-dat)
(let ((config-hostname (configf:lookup (megatest:area-configdat area-dat) "server" "hostname"))
(tdbdat (tasks:open-db area-dat)))
(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 (lambda (db portnum)
(portlogger:set-failed db area-dat portnum))
area-dat 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
(lambda (db)
(portlogger:find-port db area-dat))
area-dat)
server-id
area-dat))
(begin
(tasks:server-force-clean-run-record (db:delay-if-busy tdbdat area-dat) 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
(db:delay-if-busy tdbdat area-dat)
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 (db:delay-if-busy tdbdat area-dat) 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
;;======================================================================
;;======================================================================
|
| ︙ | | | ︙ | |
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
|
(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)
(success #t)
|
|
|
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
|
(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)(remote #f))
(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)
(success #t)
|
| ︙ | | | ︙ | |
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
|
(db:string->obj
(handle-exceptions
exn
(begin
(set! success #f)
(debug:print 0 "WARNING: failure in with-input-from-request to " fullurl ".")
(debug:print 0 " message: " ((condition-property-accessor 'exn 'message) exn))
(hash-table-delete! *runremote* run-id)
;; Killing associated server to allow clean retry.")
;; (tasks:kill-server-run-id run-id) ;; better to kill the server in the logic that called this routine?
(mutex-unlock! *http-mutex*)
;;; (signal (make-composite-condition
;;; (make-property-condition 'commfail 'message "failed to connect to server")))
;;; "communications failed"
(db:obj->string #f))
|
|
|
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
|
(db:string->obj
(handle-exceptions
exn
(begin
(set! success #f)
(debug:print 0 "WARNING: failure in with-input-from-request to " fullurl ".")
(debug:print 0 " message: " ((condition-property-accessor 'exn 'message) exn))
(common:del-remote! remote run-id)
;; Killing associated server to allow clean retry.")
;; (tasks:kill-server-run-id run-id) ;; better to kill the server in the logic that called this routine?
(mutex-unlock! *http-mutex*)
;;; (signal (make-composite-condition
;;; (make-property-condition 'commfail 'message "failed to connect to server")))
;;; "communications failed"
(db:obj->string #f))
|
| ︙ | | | ︙ | |
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
|
(pp (vector-ref res 1) (current-error-port))
(signal (vector-ref result 0))))
(signal (make-composite-condition
(make-property-condition
'timeout
'message "nmsg-transport:client-api-send-receive-raw timed out talking to server")))))))
;; careful closing of connections stored in *runremote*
;;
(define (http-transport:close-connections 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)))
|
|
|
|
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
|
(pp (vector-ref res 1) (current-error-port))
(signal (vector-ref result 0))))
(signal (make-composite-condition
(make-property-condition
'timeout
'message "nmsg-transport:client-api-send-receive-raw timed out talking to server")))))))
;; careful closing of connections stored in (common:get-remote remote)
;;
(define (http-transport:close-connections run-id)
(let* ((server-dat (common:get-remote remote run-id)))
(if (vector? server-dat)
(let ((api-dat (http-transport:server-dat-get-api-uri server-dat)))
(close-connection! api-dat)
#t)
#f)))
|
| ︙ | | | ︙ | |
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
|
(api-req (make-request method: 'POST uri: api-uri))
(server-dat (vector iface port api-uri api-url api-req (current-seconds))))
server-dat))
;; run http-transport:keep-running in a parallel thread to monitor that the db is being
;; used and to shutdown after sometime if it is not.
;;
(define (http-transport:keep-running 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
(debug:print-info 0 "Starting the sync-back, keep alive thread in server for run-id=" run-id)
(let* ((tdbdat (tasks:open-db))
(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 signature")
(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)
(if (> (- (current-seconds) start-time) 120) ;; been waiting for two minutes
(begin
(debug:print 0 "ERROR: transport appears to have died, exiting server " server-id " for run " run-id)
(tasks:server-delete-record (db:delay-if-busy tdbdat) server-id "failed to start, never received server alive signature")
(exit))
(loop start-time
(equal? sdat last-sdat)
sdat)))))))
(iface (car server-info))
(port (cadr server-info))
(last-access 0)
(server-timeout (server:get-timeout)))
(let loop ((count 0)
(server-state 'available)
(bad-sync-count 0))
;; Use this opportunity to sync the inmemdb to db
(if *inmemdb*
(let ((start-time (current-milliseconds))
(sync-time #f)
(rem-time #f))
;; inmemdb is a dbstruct
(condition-case
(db:sync-touched *inmemdb* *run-id* force-sync: #t)
((sync-failed)(cond
((> bad-sync-count 10) ;; time to give up
(http-transport:server-shutdown server-id port))
(else ;; (> bad-sync-count 0) ;; we've had a fail or two, delay and loop
(thread-sleep! 5)
(loop count server-state (+ bad-sync-count 1)))))
((exn)
(debug:print 0 "ERROR: error from sync code other than 'sync-failed. Attempting to gracefully shutdown the server")
(tasks:server-delete-record (db:delay-if-busy tdbdat) server-id " http-transport:keep-running crashed")
(exit)))
(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)
(if (and (<= rem-time 4)
(> rem-time 0))
(thread-sleep! rem-time)
(thread-sleep! 4))) ;; fallback for if the math is changed ...
;;
;; no *inmemdb* yet, set running after our first pass through and start the db
;;
(if (eq? server-state 'available)
(let ((new-server-id (tasks:server-am-i-the-server? (db:delay-if-busy tdbdat) run-id))) ;; try to ensure no double registering of servers
(if (equal? new-server-id server-id)
(begin
(tasks:server-set-state! (db:delay-if-busy tdbdat) server-id "dbprep")
(thread-sleep! 0.5) ;; give some margin for queries to complete before switching from file based access to server based access
(set! *inmemdb* (db:setup run-id))
;; force initialization
;; (db:get-db *inmemdb* #t)
(db:get-db *inmemdb* run-id)
(tasks:server-set-state! (db:delay-if-busy tdbdat) server-id "running"))
(begin ;; gotta exit nicely
(tasks:server-set-state! (db:delay-if-busy tdbdat) server-id "collision")
(http-transport:server-shutdown server-id port))))))
(if (< count 1) ;; 3x3 = 9 secs aprox
(loop (+ count 1) 'running bad-sync-count))
;; Check that iface and port have not changed (can happen if server port collides)
(mutex-lock! *heartbeat-mutex*)
(set! sdat *server-info*)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
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
|
(api-req (make-request method: 'POST uri: api-uri))
(server-dat (vector iface port api-uri api-url api-req (current-seconds))))
server-dat))
;; run http-transport:keep-running in a parallel thread to monitor that the db is being
;; used and to shutdown after sometime if it is not.
;;
(define (http-transport:keep-running server-id run-id area-dat)
;; if none running or if > 20 seconds since
;; server last used then start shutdown
;; This thread waits for the server to come alive
(debug:print-info 0 "Starting the sync-back, keep alive thread in server for run-id=" run-id)
(let* ((tdbdat (tasks:open-db area-dat))
(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 signature")
(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)
(if (> (- (current-seconds) start-time) 120) ;; been waiting for two minutes
(begin
(debug:print 0 "ERROR: transport appears to have died, exiting server " server-id " for run " run-id)
(tasks:server-delete-record (db:delay-if-busy tdbdat area-dat) server-id "failed to start, never received server alive signature")
(exit))
(loop start-time
(equal? sdat last-sdat)
sdat)))))))
(iface (car server-info))
(port (cadr server-info))
(last-access 0)
(server-timeout (server:get-timeout area-dat)))
(let loop ((count 0)
(server-state 'available)
(bad-sync-count 0))
;; Use this opportunity to sync the inmemdb to db
(if *inmemdb*
(let ((start-time (current-milliseconds))
(sync-time #f)
(rem-time #f))
;; inmemdb is a dbstruct
(condition-case
(db:sync-touched *inmemdb* area-dat *run-id* force-sync: #t)
((sync-failed)(cond
((> bad-sync-count 10) ;; time to give up
(http-transport:server-shutdown server-id port area-dat))
(else ;; (> bad-sync-count 0) ;; we've had a fail or two, delay and loop
(thread-sleep! 5)
(loop count server-state (+ bad-sync-count 1)))))
((exn)
(debug:print 0 "ERROR: error from sync code other than 'sync-failed. Attempting to gracefully shutdown the server")
(tasks:server-delete-record (db:delay-if-busy tdbdat area-dat) server-id " http-transport:keep-running crashed")
(exit)))
(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)
(if (and (<= rem-time 4)
(> rem-time 0))
(thread-sleep! rem-time)
(thread-sleep! 4))) ;; fallback for if the math is changed ...
;;
;; no *inmemdb* yet, set running after our first pass through and start the db
;;
(if (eq? server-state 'available)
(let ((new-server-id (tasks:server-am-i-the-server? (db:delay-if-busy tdbdat area-dat) run-id))) ;; try to ensure no double registering of servers
(if (equal? new-server-id server-id)
(begin
(tasks:server-set-state! (db:delay-if-busy tdbdat area-dat) server-id "dbprep")
(thread-sleep! 0.5) ;; give some margin for queries to complete before switching from file based access to server based access
(set! *inmemdb* (db:setup run-id area-dat))
;; force initialization
;; (db:get-db *inmemdb* #t)
(db:get-db *inmemdb* area-dat run-id)
(tasks:server-set-state! (db:delay-if-busy tdbdat area-dat) server-id "running"))
(begin ;; gotta exit nicely
(tasks:server-set-state! (db:delay-if-busy tdbdat area-dat) server-id "collision")
(http-transport:server-shutdown server-id port area-dat))))))
(if (< count 1) ;; 3x3 = 9 secs aprox
(loop (+ count 1) 'running bad-sync-count))
;; Check that iface and port have not changed (can happen if server port collides)
(mutex-lock! *heartbeat-mutex*)
(set! sdat *server-info*)
|
| ︙ | | | ︙ | |
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
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
|
(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*
(> (+ last-access server-timeout)
(current-seconds)))
(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 bad-sync-count))
(http-transport:server-shutdown server-id port)))))
(define (http-transport:server-shutdown server-id port)
(let ((tdbdat (tasks:open-db)))
(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! (db:delay-if-busy tdbdat) 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 (db:delay-if-busy tdbdat) server-id " http-transport:keep-running complete")
(exit)))
;; all routes though here end in exit ...
;;
;; start_server?
;;
(define (http-transport:launch run-id)
(let* ((tdbdat (tasks:open-db)))
(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 (db:delay-if-busy tdbdat) run-id))
(remtries 4))
(if (not server-id)
(if (> remtries 0)
(begin
(thread-sleep! 2)
(loop (tasks:server-lock-slot (db:delay-if-busy tdbdat) 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 (db:delay-if-busy tdbdat) " 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")))
(thread-start! th2)
(thread-sleep! 0.25) ;; give the server time to settle before starting the keep-running monitor.
(thread-start! th3)
(set! *didsomething* #t)
(thread-join! th2)
(exit))))))
(define (http:ping run-id host-port)
(let* ((server-dat (http-transport:client-connect (car host-port)(cadr host-port)))
(login-res (rmt:login-no-auto-client-setup server-dat run-id)))
(if (and (list? login-res)
(car login-res))
(begin
(print "LOGIN_OK")
(exit 0))
(begin
(print "LOGIN_FAILED")
|
|
|
|
|
|
|
|
>
>
>
|
|
|
|
|
|
|
|
>
|
|
|
|
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
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
|
(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*
(> (+ last-access server-timeout)
(current-seconds)))
(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 bad-sync-count))
(http-transport:server-shutdown server-id port area-dat)))))
(define (http-transport:server-shutdown server-id port area-dat)
(let ((tdbdat (tasks:open-db area-dat)))
(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* area-dat *run-id* force-sync: #t))
;;
;; start_shutdown
;;
(tasks:server-set-state! (db:delay-if-busy tdbdat area-dat) server-id "shutting-down")
(portlogger:open-run-close
(lambda (db port value)
(portlogger:set-port db port value))
area-dat 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 (db:delay-if-busy tdbdat area-dat) server-id " http-transport:keep-running complete")
(exit)))
;; all routes though here end in exit ...
;;
;; start_server?
;;
(define (http-transport:launch run-id area-dat)
(let* ((tdbdat (tasks:open-db area-dat)))
(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 area-dat)
(begin
(debug:print 0 "INFO: Server for run-id " run-id " already running")
(exit 0)))
(let loop ((server-id (tasks:server-lock-slot (db:delay-if-busy tdbdat area-dat) run-id area-dat))
(remtries 4))
(if (not server-id)
(if (> remtries 0)
(begin
(thread-sleep! 2)
(loop (tasks:server-lock-slot (db:delay-if-busy tdbdat area-dat) run-id area-dat)
(- 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 (db:delay-if-busy tdbdat area-dat) " 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
area-dat)) "Server run"))
(th3 (make-thread (lambda ()
(debug:print-info 0 "Server monitor thread started")
(http-transport:keep-running server-id run-id area-dat))
"Keep running")))
(thread-start! th2)
(thread-sleep! 0.25) ;; give the server time to settle before starting the keep-running monitor.
(thread-start! th3)
(set! *didsomething* #t)
(thread-join! th2)
(exit))))))
(define (http:ping run-id host-port area-dat)
(let* ((server-dat (http-transport:client-connect (car host-port)(cadr host-port)))
(login-res (rmt:login-no-auto-client-setup server-dat run-id area-dat)))
(if (and (list? login-res)
(car login-res))
(begin
(print "LOGIN_OK")
(exit 0))
(begin
(print "LOGIN_FAILED")
|
| ︙ | | | ︙ | |
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
|
(thread-start! th1)
(thread-join! th2))))
;;======================================================================
;; web pages
;;======================================================================
(define (http-transport:main-page)
(let ((linkpath (root-path)))
(conc "<head><h1>" (pathname-strip-directory *toppath*) "</h1></head>"
"Run area: " *toppath*
"<h2>Server Stats</h2>"
(http-transport:stats-table)
"<hr>"
(http-transport:runs linkpath)
"<hr>"
(http-transport:run-stats)
"</body>"
|
|
>
|
|
|
|
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
|
(thread-start! th1)
(thread-join! th2))))
;;======================================================================
;; web pages
;;======================================================================
(define (http-transport:main-page area-dat)
(let* ((toppath (megatest:area-path area-dat))
(linkpath (root-path)))
(conc "<head><h1>" (pathname-strip-directory toppath) "</h1></head>"
"<body>"
"Run area: " toppath
"<h2>Server Stats</h2>"
(http-transport:stats-table)
"<hr>"
(http-transport:runs linkpath)
"<hr>"
(http-transport:run-stats)
"</body>"
|
| ︙ | | | ︙ | |