| ︙ | | | ︙ | |
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
|
(declare (uses common))
(declare (uses db))
(declare (uses tests))
(declare (uses tasks)) ;; tasks are where stuff is maintained about what is running.
(declare (uses server))
(declare (uses daemon))
(declare (uses portlogger))
(include "common_records.scm")
(include "db_records.scm")
(define (http-transport:make-server-url hostport)
(if (not hostport)
#f
(conc "http://" (car hostport) ":" (cadr hostport))))
(define *server-loop-heart-beat* (current-seconds))
(define *heartbeat-mutex* (make-mutex))
|
>
<
|
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
|
(declare (uses common))
(declare (uses db))
(declare (uses tests))
(declare (uses tasks)) ;; tasks are where stuff is maintained about what is running.
(declare (uses server))
(declare (uses daemon))
(declare (uses portlogger))
(declare (uses rmt))
(include "common_records.scm")
(include "db_records.scm")
(define (http-transport:make-server-url hostport)
(if (not hostport)
#f
(conc "http://" (car hostport) ":" (cadr hostport))))
(define *server-loop-heart-beat* (current-seconds))
;;======================================================================
;; S E R V E R
;;======================================================================
;; Call this to start the actual server
;;
|
| ︙ | | | ︙ | |
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
|
(ipaddrstr (let ((ipstr (if (string=? "-" hostn)
;; (string-intersperse (map number->string (u8vector->list (hostname->ip hostname))) ".")
(server:get-best-guess-address hostname)
#f)))
(if ipstr ipstr hostn))) ;; hostname)))
(start-port (portlogger:open-run-close portlogger:find-port))
(link-tree-path (configf:lookup *configdat* "setup" "linktree")))
;; (set! db *inmemdb*)
|
<
|
56
57
58
59
60
61
62
63
64
65
66
67
68
69
|
(ipaddrstr (let ((ipstr (if (string=? "-" hostn)
;; (string-intersperse (map number->string (u8vector->list (hostname->ip hostname))) ".")
(server:get-best-guess-address hostname)
#f)))
(if ipstr ipstr hostn))) ;; hostname)))
(start-port (portlogger:open-run-close portlogger:find-port))
(link-tree-path (configf:lookup *configdat* "setup" "linktree")))
(debug:print-info 0 *default-log-port* "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)
(signal (make-composite-condition
|
| ︙ | | | ︙ | |
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
|
;; 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)))
'(/ "json_api"))
(send-response body: (http-transport:main-page)))
|
|
|
|
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
|
;; This is were we set up the database connections
(let* (($ (request-vars source: 'both))
(dat ($ 'dat))
(res #f))
(cond
((equal? (uri-path (request-uri (current-request)))
'(/ "api"))
(send-response body: (api:process-request *dbstruct-db* $) ;; the $ is the request vars proc
headers: '((content-type text/plain)))
(mutex-lock! *heartbeat-mutex*)
(set! *db-last-access* (current-seconds))
(mutex-unlock! *heartbeat-mutex*))
((equal? (uri-path (request-uri (current-request)))
'(/ ""))
(send-response body: (http-transport:main-page)))
((equal? (uri-path (request-uri (current-request)))
'(/ "json_api"))
(send-response body: (http-transport:main-page)))
|
| ︙ | | | ︙ | |
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
|
(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 *default-log-port* "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 *default-log-port* "WARNING: attempt to start server failed. Trying again ...")
|
|
|
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
|
(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 *default-log-port* "http-transport:try-start-server time=" (seconds->time-string (current-seconds)) " 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 *default-log-port* "WARNING: attempt to start server failed. Trying again ...")
|
| ︙ | | | ︙ | |
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
|
(http-transport:server-dat-get-api-req serverdat)
(begin
(debug:print 0 *default-log-port* "FATAL ERROR: http-transport:client-api-send-receive called with no server info")
(exit 1))))
(res #f)
(success #t)
(sparams (db:obj->string params transport: 'http)))
;; (condition-case
;; handle-exceptions
;; exn
;; (if (> numretries 0)
;; (begin
;; (mutex-unlock! *http-mutex*)
;; (thread-sleep! 1)
;; (handle-exceptions
;; exn
;; (debug:print 0 *default-log-port* "WARNING: closing connections failed. Server at " fullurl " almost certainly dead")
;; (close-all-connections!))
;; (debug:print 0 *default-log-port* "WARNING: Failed to communicate with server, trying again, numretries left: " numretries)
;; (http-transport:client-api-send-receive run-id serverdat cmd sparams numretries: (- numretries 1)))
;; (begin
;; (mutex-unlock! *http-mutex*)
;; (tasks:kill-server-run-id run-id)
;; #f))
;; (begin
|
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
|
218
219
220
221
222
223
224
225
226
227
228
229
230
231
|
(http-transport:server-dat-get-api-req serverdat)
(begin
(debug:print 0 *default-log-port* "FATAL ERROR: http-transport:client-api-send-receive called with no server info")
(exit 1))))
(res #f)
(success #t)
(sparams (db:obj->string params transport: 'http)))
(debug:print-info 11 *default-log-port* "fullurl=" fullurl ", cmd=" cmd ", params=" params ", run-id=" run-id "\n")
;; set up the http-client here
(max-retry-attempts 1)
;; consider all requests indempotent
(retry-request? (lambda (request)
#f))
;; send the data and get the response
|
| ︙ | | | ︙ | |
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
|
(db:string->obj
(handle-exceptions
exn
(begin
(set! success #f)
(debug:print 0 *default-log-port* "WARNING: failure in with-input-from-request to " fullurl ".")
(debug:print 0 *default-log-port* " 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))
(with-input-from-request ;; was dat
fullurl
(list (cons 'key "thekey")
(cons 'cmd cmd)
(cons 'params sparams))
read-string))
transport: 'http)))
;; Shouldn't this be a call to the managed call-all-connections stuff above?
(close-all-connections!)
(mutex-unlock! *http-mutex*)
))
(time-out (lambda ()
(thread-sleep! 45)
#f))
(th1 (make-thread send-recieve "with-input-from-request"))
(th2 (make-thread time-out "time out")))
(thread-start! th1)
(thread-start! th2)
(thread-join! th1)
(thread-terminate! th2)
(debug:print-info 11 *default-log-port* "got res=" res)
(if (vector? res)
(if (vector-ref res 0)
res
(begin ;; note: this code also called in nmsg-transport - consider consolidating it
(debug:print-error 0 *default-log-port* "error occured at server, info=" (vector-ref res 2))
(debug:print 0 *default-log-port* " client call chain:")
(print-call-chain (current-error-port))
(debug:print 0 *default-log-port* " server call chain:")
(pp (vector-ref res 1) (current-error-port))
(signal (vector-ref result 0))))
(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)))
|
|
>
|
>
>
|
|
|
|
|
|
|
>
>
>
|
|
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
|
(db:string->obj
(handle-exceptions
exn
(begin
(set! success #f)
(debug:print 0 *default-log-port* "WARNING: failure in with-input-from-request to " fullurl ".")
(debug:print 0 *default-log-port* " message: " ((condition-property-accessor 'exn 'message) exn))
(if *runremote*
(remote-conndat-set! *runremote* #f))
;; Killing associated server to allow clean retry.")
;; (tasks:kill-server-run-id run-id) ;; better to kill the server in the logic that called this routine?
(mutex-unlock! *http-mutex*)
;;; (signal (make-composite-condition
;;; (make-property-condition 'commfail 'message "failed to connect to server")))
;;; "communications failed"
(db:obj->string #f))
(with-input-from-request ;; was dat
fullurl
(list (cons 'key "thekey")
(cons 'cmd cmd)
(cons 'params sparams))
read-string))
transport: 'http)
0)) ;; added this speculatively
;; Shouldn't this be a call to the managed call-all-connections stuff above?
(close-all-connections!)
(mutex-unlock! *http-mutex*)
))
(time-out (lambda ()
(thread-sleep! 45)
#f))
(th1 (make-thread send-recieve "with-input-from-request"))
(th2 (make-thread time-out "time out")))
(thread-start! th1)
(thread-start! th2)
(thread-join! th1)
(thread-terminate! th2)
(debug:print-info 11 *default-log-port* "got res=" res)
(if (vector? res)
(if (vector-ref res 0)
res
(if (debug:debug-mode 11)
(begin ;; note: this code also called in nmsg-transport - consider consolidating it
(debug:print-error 11 *default-log-port* "error occured at server, info=" (vector-ref res 2))
(debug:print 11 *default-log-port* " client call chain:")
(print-call-chain (current-error-port))
(debug:print 11 *default-log-port* " server call chain:")
(pp (vector-ref res 1) (current-error-port))
(signal (vector-ref res 0)))
res))
(signal (make-composite-condition
(make-property-condition
'timeout
'message "nmsg-transport:client-api-send-receive-raw timed out talking to server")))))))
;; careful closing of connections stored in *runremote*
;;
(define (http-transport:close-connections run-id)
(let* ((server-dat (if *runremote*
(remote-conndat *runremote*)
#f))) ;; (hash-table-ref/default *runremote* run-id #f)))
(if (vector? server-dat)
(let ((api-dat (http-transport:server-dat-get-api-uri server-dat)))
(close-connection! api-dat)
#t)
#f)))
|
| ︙ | | | ︙ | |
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
|
(debug:print-info 0 *default-log-port* "Waiting for server alive signature")
(mutex-lock! *heartbeat-mutex*)
(set! sdat *server-info*)
(mutex-unlock! *heartbeat-mutex*)
(if (and sdat
(not changed)
(> (- (current-seconds) start-time) 2))
sdat
(begin
(debug:print-info 0 *default-log-port* "Still waiting, last-sdat=" last-sdat)
(sleep 4)
(if (> (- (current-seconds) start-time) 120) ;; been waiting for two minutes
(begin
(debug:print-error 0 *default-log-port* "transport appears to have died, exiting server " 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-error 0 *default-log-port* "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 4 *default-log-port* "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*)
(mutex-unlock! *heartbeat-mutex*)
(if (or (not (equal? sdat (list iface port)))
(not server-id))
(begin
(debug:print-info 0 *default-log-port* "interface changed, refreshing iface and port info")
(set! iface (car sdat))
(set! port (cadr sdat))))
;; Transfer *last-db-access* to last-access to use in checking that we are still alive
(mutex-lock! *heartbeat-mutex*)
(set! last-access *last-db-access*)
(mutex-unlock! *heartbeat-mutex*)
;; (debug:print 11 *default-log-port* "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" b"wait-on-running"))) ;; wait on running tasks (if not true then exit on time out)
|
>
>
|
|
>
|
>
|
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
|
<
<
<
<
|
<
|
<
|
|
|
|
|
|
|
|
<
<
|
>
>
|
|
|
>
>
>
>
>
>
>
>
|
|
|
|
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
|
(debug:print-info 0 *default-log-port* "Waiting for server alive signature")
(mutex-lock! *heartbeat-mutex*)
(set! sdat *server-info*)
(mutex-unlock! *heartbeat-mutex*)
(if (and sdat
(not changed)
(> (- (current-seconds) start-time) 2))
(begin
(debug:print-info 0 *default-log-port* "Received server alive signature")
sdat)
(begin
(debug:print-info 0 *default-log-port* "Still waiting, last-sdat=" last-sdat)
(sleep 4)
(if (> (- (current-seconds) start-time) 120) ;; been waiting for two minutes
(begin
(debug:print-error 0 *default-log-port* "transport appears to have died, exiting server " 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))
(server-going #f))
(let loop ((count 0)
(server-state 'available)
(bad-sync-count 0)
(start-time (current-milliseconds)))
;; Use this opportunity to sync the tmp db to megatest.db
(if (not server-going) ;; *dbstruct-db*
;; Removed code is pasted below (keeping it around until we are clear it is not needed).
;; no *dbstruct-db* 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! *dbstruct-db* (db:setup)) ;; run-id))
(set! server-going #t)
(tasks:server-set-state! (db:delay-if-busy tdbdat) server-id "running")
(server:write-dotserver *toppath* (conc iface ":" port))
(delete-file* (conc *toppath* "/.starting-server")))
(begin ;; gotta exit nicely
(tasks:server-set-state! (db:delay-if-busy tdbdat) server-id "collision")
(http-transport:server-shutdown server-id port))))))
;; when things go wrong we don't want to be doing the various queries too often
;; so we strive to run this stuff only every four seconds or so.
(let* ((sync-time (- (current-milliseconds) start-time))
(rem-time (quotient (- 4000 sync-time) 1000)))
(if (and (<= rem-time 4)
(> rem-time 0))
(thread-sleep! rem-time)))
(if (< count 1) ;; 3x3 = 9 secs aprox
(loop (+ count 1) 'running bad-sync-count (current-milliseconds)))
;; Check that iface and port have not changed (can happen if server port collides)
(mutex-lock! *heartbeat-mutex*)
(set! sdat *server-info*)
(mutex-unlock! *heartbeat-mutex*)
(if (or (not (equal? sdat (list iface port)))
(not server-id))
(begin
(debug:print-info 0 *default-log-port* "interface changed, refreshing iface and port info")
(set! iface (car sdat))
(set! port (cadr sdat))))
;; Transfer *db-last-access* to last-access to use in checking that we are still alive
(mutex-lock! *heartbeat-mutex*)
(set! last-access *db-last-access*)
(mutex-unlock! *heartbeat-mutex*)
;; (debug:print 11 *default-log-port* "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" b"wait-on-running"))) ;; wait on running tasks (if not true then exit on time out)
|
| ︙ | | | ︙ | |
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
|
;;
;; 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))))))
;; 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))
(if (server:check-if-running run-id)
(begin
(debug:print 0 *default-log-port* "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 *default-log-port* "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 *default-log-port* "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 *default-log-port* "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")
(exit 1)))))
(define (http-transport:server-signal-handler signum)
(signal-mask! signum)
(handle-exceptions
exn
(debug:print 0 *default-log-port* " ... exiting ...")
(let ((th1 (make-thread (lambda ()
|
|
|
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
<
<
<
>
>
>
>
>
>
>
>
|
|
>
>
>
|
|
|
|
|
|
|
|
|
|
|
|
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
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
|
;;
;; 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 (current-milliseconds)))
(http-transport:server-shutdown server-id port))))))
;; code cut out from above
;;
;; (condition-case
;; ;; (if (and (member (mutex-state *db-sync-mutex*) '(abandoned not-abandoned))
;; ;; (> (- (current-seconds) *db-last-sync*) 5)) ;; if not currently being synced nor recently synced
;; (db:sync-touched *dbstruct-db* *run-id* force-sync: #t) ;; usually done in the watchdog, not here.
;; ((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-error 0 *default-log-port* "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 4 *default-log-port* "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 ...
(define (http-transport:server-shutdown server-id port)
(let ((tdbdat (tasks:open-db)))
(debug:print-info 0 *default-log-port* "Starting to shutdown the server.")
;;
;; start_shutdown
;;
(tasks:server-set-state! (db:delay-if-busy tdbdat) server-id "shutting-down")
(set! *time-to-exit* #t) ;; tell on-exit to be fast as we've already cleaned up
(portlogger:open-run-close portlogger:set-port port "released")
(thread-sleep! 5)
(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")
(debug:print-info 0 *default-log-port* "Server shutdown complete. Exiting")
(tasks:server-delete-record (db:delay-if-busy tdbdat) server-id " http-transport:keep-running complete")
;; if the .server file contained :myport then we can remove it
(server:remove-dotserver-file *toppath* port)
(exit)))
;; all routes though here end in exit ...
;;
;; start_server?
;;
(define (http-transport:launch run-id)
(with-output-to-file
(conc *toppath* "/.starting-server")
(lambda ()
(print (current-process-id) " on " (get-host-name))))
(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 (and (server:read-dotserver *toppath*)
(server:check-if-running run-id))
(begin
(debug:print 0 *default-log-port* "INFO: Server for run-id " run-id " already running")
(exit 0))
(begin ;; ok, no server detected, clean out any lingering records
(tasks:server-force-clean-running-records-for-run-id (db:delay-if-busy tdbdat) run-id "notresponding")))
(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 *default-log-port* "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")
(delete-file* (conc *toppath* "/.starting-server"))
))
(let* ((th2 (make-thread (lambda ()
(debug:print-info 0 *default-log-port* "Server run thread started")
(http-transport:run
(if (args:get-arg "-server")
(args:get-arg "-server")
"-")
run-id
server-id)) "Server run"))
(th3 (make-thread (lambda ()
(debug:print-info 0 *default-log-port* "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")
;; (exit 1)))))
(define (http-transport:server-signal-handler signum)
(signal-mask! signum)
(handle-exceptions
exn
(debug:print 0 *default-log-port* " ... exiting ...")
(let ((th1 (make-thread (lambda ()
|
| ︙ | | | ︙ | |
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
|
" ms</td></tr>"
"<tr><td>Number non-cached queries</td> <td>" *number-non-write-queries* "</td></tr>"
"<tr><td>Average non-cached time</td> <td>" (if (eq? *number-non-write-queries* 0)
"n/a (no queries)"
(/ *total-non-write-delay*
*number-non-write-queries*))
" ms</td></tr>"
"<tr><td>Last access</td><td>" (seconds->time-string *last-db-access*) "</td></tr>"
"</table>")))
(mutex-unlock! *heartbeat-mutex*)
res))
(define (http-transport:runs linkpath)
(conc "<h3>Runs</h3>"
(string-intersperse
|
|
|
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
|
" ms</td></tr>"
"<tr><td>Number non-cached queries</td> <td>" *number-non-write-queries* "</td></tr>"
"<tr><td>Average non-cached time</td> <td>" (if (eq? *number-non-write-queries* 0)
"n/a (no queries)"
(/ *total-non-write-delay*
*number-non-write-queries*))
" ms</td></tr>"
"<tr><td>Last access</td><td>" (seconds->time-string *db-last-access*) "</td></tr>"
"</table>")))
(mutex-unlock! *heartbeat-mutex*)
res))
(define (http-transport:runs linkpath)
(conc "<h3>Runs</h3>"
(string-intersperse
|
| ︙ | | | ︙ | |