| ︙ | | |
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
|
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
|
-
+
|
;; (string-intersperse (map number->string (u8vector->list (hostname->ip hostname))) ".")
(server:get-best-guess-address hostname)
#f)))
(if ipstr ipstr hostn))) ;; hostname)))
(start-port (portlogger:open-run-close portlogger:find-port))
(link-tree-path (configf:lookup *configdat* "setup" "linktree")))
;; (set! db *inmemdb*)
(debug:print-info 0 #f "portlogger recommended port: " start-port)
(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
(make-property-condition
|
| ︙ | | |
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
|
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 #f "http-transport:try-start-server run-id=" run-id " ipaddrsstr=" ipaddrstr " portnum=" portnum " server-id=" server-id " config-hostname=" config-hostname)
(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 ...")
|
| ︙ | | |
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
|
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
|
-
+
|
(define (http-transport:inc-requests-count)
(mutex-lock! *http-mutex*)
(set! *http-requests-in-progress* (+ 1 *http-requests-in-progress*))
;; Use this opportunity to slow things down iff there are too many requests in flight
(if (> *http-requests-in-progress* 5)
(begin
(debug:print-info 0 #f "Whoa there buddy, ease up...")
(debug:print-info 0 *default-log-port* "Whoa there buddy, ease up...")
(thread-sleep! 1)))
(mutex-unlock! *http-mutex*))
(define (http-transport:dec-requests-count proc)
(mutex-lock! *http-mutex*)
(proc)
(set! *http-requests-in-progress* (- *http-requests-in-progress* 1))
|
| ︙ | | |
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
|
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
|
-
+
|
;; (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
(debug:print-info 11 #f "fullurl=" fullurl ", cmd=" cmd ", params=" params ", run-id=" run-id "\n")
(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
;; extract the needed info from the http data and
|
| ︙ | | |
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
|
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
|
-
+
|
#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 #f "got res=" res)
(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 0 *default-log-port* "ERROR: error occured at server, info=" (vector-ref res 2))
(debug:print 0 *default-log-port* " client call chain:")
(print-call-chain (current-error-port))
|
| ︙ | | |
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
|
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
|
-
+
-
+
-
+
|
;; 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 #f "Starting the sync-back, keep alive thread in server for run-id=" run-id)
(debug:print-info 0 *default-log-port* "Starting the sync-back, keep alive thread in server for run-id=" run-id)
(let* ((tdbdat (tasks:open-db))
(server-start-time (current-seconds))
(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 #f "Waiting for server alive signature")
(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 #f "Still waiting, last-sdat=" last-sdat)
(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 0 *default-log-port* "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
|
| ︙ | | |
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
|
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
|
-
+
-
+
-
+
-
+
-
-
-
+
+
+
-
-
+
+
-
+
|
(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 #f "interface changed, refreshing iface and port info")
(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)
;;
(let* ((hrs-since-start (/ (- (current-seconds) server-start-time) 3600))
(adjusted-timeout (if (> hrs-since-start 1)
(- server-timeout (inexact->exact (round (* hrs-since-start 60)))) ;; subtract 60 seconds per hour
server-timeout)))
(if (common:low-noise-print 120 "server timeout")
(debug:print-info 0 #f "Adjusted server timeout: " adjusted-timeout))
(debug:print-info 0 *default-log-port* "Adjusted server timeout: " adjusted-timeout))
(if (and *server-run*
(> (+ last-access server-timeout)
(current-seconds)))
(begin
(if (common:low-noise-print 120 "server continuing")
(debug:print-info 0 #f "Server continuing, seconds since last db access: " (- (current-seconds) last-access)))
(debug:print-info 0 *default-log-port* "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 #f "Starting to shutdown the server.")
(debug:print-info 0 *default-log-port* "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 #f "Max cached queries was " *max-cache-size*)
(debug:print-info 0 #f "Number of cached writes " *number-of-writes*)
(debug:print-info 0 #f "Average cached write time "
(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 #f "Number non-cached queries " *number-non-write-queries*)
(debug:print-info 0 #f "Average non-cached time "
(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 #f "Server shutdown complete. Exiting")
(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")
(exit)))
;; all routes though here end in exit ...
;;
;; start_server?
;;
|
| ︙ | | |
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
|
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
|
-
+
-
+
-
+
|
(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 #f "INFO: server pid=" (current-process-id) ", hostname=" (get-host-name) " not starting due to other candidates ahead in start queue")
(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 #f "Server run thread started")
(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 #f "Server monitor thread started")
(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)
|
| ︙ | | |