︙ | | |
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
|
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
|
-
+
-
-
-
-
-
-
-
-
-
-
-
-
|
;;(declare (uses rpc-transport))
(declare (uses launch))
;; (declare (uses daemon))
(include "common_records.scm")
(include "db_records.scm")
(define (server:make-server-url hostport)
#;(define (server:make-server-url hostport)
(if (not hostport)
#f
(conc "http://" (car hostport) ":" (cadr hostport))))
(define *server-loop-heart-beat* (current-seconds))
;;======================================================================
;; P K T S S T U F F
;;======================================================================
;; ???
;;======================================================================
;; P K T S S T U F F
;;======================================================================
;; ???
;;======================================================================
;; S E R V E R
;;======================================================================
;; Call this to start the actual server
;;
|
︙ | | |
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
|
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
|
-
+
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
|
(else (debug:print-error 0 *default-log-port* "unknown server type " transport-type))))
;;======================================================================
;; S E R V E R U T I L I T I E S
;;======================================================================
;; Get the transport
(define (server:get-transport)
#;(define (server:get-transport)
(if *transport-type*
*transport-type*
(let ((ttype (string->symbol
(or (args:get-arg "-transport")
(configf:lookup *configdat* "server" "transport")
"rpc"))))
(set! *transport-type* ttype)
ttype)))
;; Generate a unique signature for this server
(define (server:mk-signature)
(message-digest-string (md5-primitive)
(with-output-to-string
(lambda ()
(write (list (current-directory)
(argv)))))))
;; When using zmq this would send the message back (two step process)
;; with spiffy or rpc this simply returns the return data to be returned
;;
(define (server:reply return-addr query-sig success/fail result)
(debug:print-info 11 *default-log-port* "server:reply return-addr=" return-addr ", result=" result)
;; (send-message pubsock target send-more: #t)
;; (send-message pubsock
(case (server:get-transport)
((rpc) (db:obj->string (vector success/fail query-sig result)))
((http) (db:obj->string (vector success/fail query-sig result)))
((fs) result)
(else
(debug:print-error 0 *default-log-port* "unrecognised transport type: " *transport-type*)
result)))
;; Given a run id start a server process ### NOTE ### > file 2>&1
;; if the run-id is zero and the target-host is set
;; try running on that host
;; incidental: rotate logs in logs/ dir.
;;
(define (server:run areapath) ;; areapath is *toppath* for a given testsuite area
(let* ((curr-host (get-host-name))
|
︙ | | |
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
|
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
|
-
+
|
(define (server:record->url servr)
(match-let (((mod-time host port start-time pid)
servr))
(if (and host port)
(conc host ":" port)
#f)))
(define (server:get-client-signature) ;; BB> why is this proc named "get-"? it returns nothing -- set! has not return value.
#;(define (server:get-client-signature) ;; BB> why is this proc named "get-"? it returns nothing -- set! has not return value.
(if *my-client-signature* *my-client-signature*
(let ((sig (server:mk-signature)))
(set! *my-client-signature* sig)
*my-client-signature*)))
;; wait for server=start-last to be three seconds old
;;
|
︙ | | |
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
|
362
363
364
365
366
367
368
369
370
371
372
373
374
375
|
-
-
|
(if (and (> try-num 0) ;; first time through simply wait a little while then try again
(< num-ok 1)) ;; if there are no decent candidates for servers then try starting a new one
(server:kind-run areapath))
(thread-sleep! 5)
(loop (server:check-if-running areapath)
(+ try-num 1)))))))
(define server:try-running server:run) ;; there is no more per-run servers ;; REMOVE ME. BUG.
(define (server:get-num-servers #!key (numservers 2))
(let ((ns (string->number
(or (configf:lookup *configdat* "server" "numservers") "notanumber"))))
(or ns numservers)))
;; no longer care if multiple servers are started by accident. older servers will drop off in time.
;;
|
︙ | | |
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
|
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
|
-
-
-
-
-
-
-
+
-
-
|
;;
;; NOTE: This is NOT called directly from clients as not all transports support a client running
;; in the same process as the server.
;;
(define (server:ping host-port-in #!key (do-exit #f))
(let ((host:port (if (not host-port-in) ;; use read-dotserver to find
#f ;; (server:check-if-running *toppath*)
;; (if (number? host-port-in) ;; we were handed a server-id
;; (let ((srec (tasks:get-server-by-id (db:delay-if-busy (tasks:open-db)) host-port-in)))
;; ;; (print "srec: " srec " host-port-in: " host-port-in)
;; (if srec
;; (conc (vector-ref srec 3) ":" (vector-ref srec 4))
;; (conc "no such server-id " host-port-in)))
host-port-in))) ;; )
host-port-in)))
(let* ((host-port (if host:port
(let ((slst (string-split host:port ":")))
(if (eq? (length slst) 2)
(list (car slst)(string->number (cadr slst)))
#f))
#f)))
;; (toppath (launch:setup)))
;; (print "host-port=" host-port)
(if (not host-port)
(begin
(if host-port-in
(debug:print 0 *default-log-port* "ERROR: bad host:port"))
(if do-exit (exit 1))
#f)
(let* ((iface (car host-port))
|
︙ | | |
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
|
456
457
458
459
460
461
462
463
464
465
466
467
468
469
|
-
-
-
-
-
-
-
-
-
|
(if (eof-object? inl)
(case (string->symbol res)
((NOREPLY) #f)
((LOGIN_OK) #t)
(else #f))
(loop (read-line) inl))))))
;; NOT USED (well, ok, reference in rpc-transport but otherwise not used).
;;
(define (server:login toppath)
(lambda (toppath)
(set! *db-last-access* (current-seconds)) ;; might not be needed.
(if (equal? *toppath* toppath)
#t
#f)))
;; timeout is hms string: 1h 5m 3s, default is 1 minute
;;
(define (server:expiration-timeout)
(let ((tmo (configf:lookup *configdat* "server" "timeout")))
(if (and (string? tmo)
(common:hms-string->seconds tmo)) ;; BUG: hms-string->seconds is broken, if given "10" returns 0. Also, it doesn't belong in this logic unless the string->number is changed below
(* 3600 (string->number tmo))
|
︙ | | |
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
|
478
479
480
481
482
483
484
485
486
487
488
489
490
491
|
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
|
;; 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 server:sync-lock-token "SERVER_SYNC_LOCK")
;; (define (server:release-sync-lock)
;; (db:no-sync-del! *no-sync-db* server:sync-lock-token))
;; (define (server:have-sync-lock?)
;; (let* ((have-lock-pair (db:no-sync-get-lock *no-sync-db* server:sync-lock-token))
;; (have-lock? (car have-lock-pair))
;; (lock-time (cdr have-lock-pair))
;; (lock-age (- (current-seconds) lock-time)))
;; (cond
;; (have-lock? #t)
;; ((>lock-age
;; (* 3 (configf:lookup-number *configdat* "server" "minimum-intersync-delay" default: 180)))
;; (server:release-sync-lock)
;; (server:have-sync-lock?))
;; (else #f))))
;; moving this here as it needs access to db and cannot be in common.
;;
(define (server:get-bruteforce-syncer dbstruct #!key (fork-to-background #f) (persist-until-sync #f))
(let* ((sqlite-exe (or (get-environment-variable "MT_SQLITE3_EXE"))) ;; defined in cfg.sh
(sync-log (or (args:get-arg "-sync-log") (conc *toppath* "/logs/sync-" (current-process-id) "-" (get-host-name) ".log")))
(tmp-area (common:get-db-tmp-area))
|
︙ | | |
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
|
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
|
-
+
|
(duty-cycle (configf:lookup-number *configdat* "server" "sync-duty-cycle" default: default-duty-cycle))
(last-sync-seconds 10) ;; we will adjust this to a measurement and delay last-sync-seconds * (1 - duty-cycle)
(calculate-off-time (lambda (work-duration duty-cycle)
(* (/ (- 1 duty-cycle) duty-cycle) last-sync-seconds)))
(off-time min-intersync-delay) ;; adjusted in closure below.
(do-a-sync
(lambda ()
(BB> "Start do-a-sync with fork-to-background="fork-to-background" persist-until-sync="persist-until-sync)
;; (BB> "Start do-a-sync with fork-to-background="fork-to-background" persist-until-sync="persist-until-sync)
(let* ((finalres
(let retry-loop ((num-tries 0))
(if (common:simple-file-lock lockfile)
(begin
(cond
((not (or fork-to-background persist-until-sync))
(debug:print 0 *default-log-port* "INFO: syncer thread sleeping for max of (server.minimum-intersync-delay="min-intersync-delay
|
︙ | | |
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
|
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
|
-
-
-
-
|
(else
(system (conc "/bin/cp "sync-log" "sync-log".fail"))
(debug:print 0 *default-log-port* "ERROR: ["(common:human-time)"] Sync failed. See log at "sync-log".fail")
(if (file-exists? (conc mtdbfile ".backup"))
(system (conc "/bin/cp "mtdbfile ".backup " mtdbfile)))
#f))))
(common:simple-file-release-lock lockfile)
(BB> "released lockfile: " lockfile)
(when (common:file-exists? lockfile)
(BB> "DID NOT ACTUALLY RELEASE LOCKFILE"))
res2) ;; end let
);; end begin
;; else
(cond
(persist-until-sync
(thread-sleep! 1)
(debug:print 1 *default-log-port* "INFO: ["(common:human-time)"] pid="(current-process-id)" other SYNC in progress; we're in a fork-to-background so we need to succeed. Let's wait a jiffy and and try again. num-tries="num-tries" (waiting for lockfile="lockfile" to disappear)")
(retry-loop (add1 num-tries)))
(else
(thread-sleep! (max off-time (+ last-sync-seconds min-intersync-delay)))
(debug:print 1 *default-log-port* "INFO: ["(common:human-time)"] pid="(current-process-id)" other SYNC in progress; not syncing.")
'parallel-sync-in-progress))
) ;; end if got lockfile
)
))
(BB> "End do-a-sync with fork-to-background="fork-to-background" persist-until-sync="persist-until-sync" and result="finalres)
finalres)
) ;; end lambda
))
do-a-sync))
(define (server:writable-watchdog-bruteforce dbstruct)
(thread-sleep! 1) ;; delay for startup
|
︙ | | |
731
732
733
734
735
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
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
|
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
|
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
|
(if (> res 0) ;; some records were transferred, keep the db alive
(begin
(mutex-lock! *heartbeat-mutex*)
(set! *db-last-access* (current-seconds))
(mutex-unlock! *heartbeat-mutex*)
(debug:print-info 0 *default-log-port* "sync called, " res " records transferred."))
(debug:print-info 2 *default-log-port* "sync called but zero records transferred")))))
;; ;; TODO: factor this next routine out into a function
;; (with-input-from-pipe ;; this should not block other threads but need to verify this
;; (conc "megatest -sync-to-megatest.db -m testsuite:" (common:get-area-name) ":" *toppath*)
;; (lambda ()
;; (let loop ((inl (read-line))
;; (res #f))
;; (if (eof-object? inl)
;; (begin
;; (set! sync-duration (- (current-milliseconds) sync-start))
;; (cond
;; ((not res)
;; (debug:print 0 *default-log-port* "ERROR: sync from /tmp db to megatest.db appears to have failed. Recommended that you stop your runs and run \"megatest -cleanup-db\""))
;; ((> res 0)
;; (mutex-lock! *heartbeat-mutex*)
;; (set! *db-last-access* (current-seconds))
;; (mutex-unlock! *heartbeat-mutex*))))
;; (let ((num-synced (let ((matches (string-match "^Synced (\\d+).*$" inl)))
;; (if matches
;; (string->number (cadr matches))
;; #f))))
;; (loop (read-line)
;; (or num-synced res))))))))))
(if will-sync
(begin
(mutex-lock! *db-multi-sync-mutex*)
(set! *db-sync-in-progress* #f)
(set! *db-last-sync* start-time)
(with-output-to-file end-file (lambda ()(print (current-process-id))))
;; release lock here
(mutex-unlock! *db-multi-sync-mutex*)))
(if (and debug-mode
(> (- start-time last-time) 60))
(begin
(set! last-time start-time)
(debug:print-info 4 *default-log-port* "timestamp -> " (seconds->time-string (current-seconds)) ", time since start -> " (seconds->hr-min-sec (- (current-seconds) *time-zero*))))))
;; keep going unless time to exit
;;
(if (not *time-to-exit*)
(let delay-loop ((count 0))
;;(debug:print-info 13 *default-log-port* "delay-loop top; count="count" pid="(current-process-id)" this-wd-num="this-wd-num" *time-to-exit*="*time-to-exit*)
(if (and (not *time-to-exit*)
(< count 6)) ;; was 11, changing to 4.
(begin
(thread-sleep! 1)
(delay-loop (+ count 1))))
(if (not *time-to-exit*) (loop))))
;; time to exit, close the no-sync db here
(db:no-sync-close-db no-sync-db stmt-cache)
(if (common:low-noise-print 30)
(debug:print-info 0 *default-log-port* "Exiting watchdog timer, *time-to-exit* = " *time-to-exit*" pid="(current-process-id) ))))))) ;;" this-wd-num="this-wd-num)))))))
|