326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
|
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
|
-
+
|
(else 600))
(random 5))) ;; add a small random number just in case a lot of jobs hit the work hosts simultaneously
(lock-file (conc areapath "/logs/server-start.lock")))
(if (> (- (current-seconds) when-run) run-delay)
(begin
(common:simple-file-lock-and-wait lock-file expire-time: 15)
(server:run areapath)
(thread-sleep! 5) ;; don't release the lock for at least a few seconds
(thread-sleep! 2) ;; don't release the lock for at least a few seconds
(common:simple-file-release-lock lock-file)))
(hash-table-set! *server-kind-run* areapath (list (+ call-num 1)(current-seconds))))))
(define (server:start-and-wait areapath #!key (timeout 60))
(let ((give-up-time (+ (current-seconds) timeout)))
(let loop ((server-url (server:check-if-running areapath))
(try-num 0))
|
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
597
598
599
600
601
602
603
604
605
606
607
608
609
610
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
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
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
703
704
705
706
|
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
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
|
-
-
+
+
-
-
+
-
-
-
+
+
+
+
+
+
+
-
-
-
+
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
-
-
-
-
-
-
-
-
-
-
-
-
-
+
+
+
-
-
-
-
-
+
-
-
-
-
-
+
+
+
+
-
-
-
-
-
-
-
+
-
-
-
-
-
-
-
-
-
-
+
-
-
-
-
-
+
+
+
-
-
-
-
-
-
-
-
-
+
+
+
+
+
+
+
+
+
+
+
+
+
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
+
+
+
+
+
+
-
-
-
-
-
-
+
+
-
-
-
-
-
+
+
-
-
-
-
-
-
+
+
+
+
+
+
+
+
+
-
-
-
+
-
-
-
-
-
-
-
-
+
+
+
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
+
-
-
-
-
-
-
-
-
-
+
+
-
-
-
-
-
-
-
-
-
+
-
-
+
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
+
+
-
-
-
-
-
-
-
-
-
-
-
-
-
-
+
+
+
+
+
+
+
|
;; (* 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:writable-watchdog dbstruct)
(thread-sleep! 10) ;; delay for startup
(let* ((legacy-sync (common:run-sync?))
(sqlite-exe (or (get-environment-variable "MT_SQLITE3_EXE"))) ;; defined in cfg.sh
(lockfile (conc tmp-db ".lock"))
(sync-cmd (conc sqlite-exe" " tmp-db " .dump | "sqlite-exe" " staging-file "&>"sync-log))
(min-intersync-delay (configf:lookup-number *configdat* "server" "minimum-intersync-delay" default: 30)))
(lockfile (common:get-sync-lock-filepath))
(sync-cmd-core (conc sqlite-exe" " tmp-db " .dump | "sqlite-exe" " staging-file "&>"sync-log))
(sync-cmd (if fork-to-background
(conc "/usr/bin/env NBFAKE_LOG="*toppath*"/logs/last-server-sync-"(current-process-id)".log nbfake \""sync-cmd-core" && /bin/mv -f " staging-file " " mtdbfile" \"")
sync-cmd-core))
(default-min-intersync-delay 2)
(min-intersync-delay (configf:lookup-number *configdat* "server" "minimum-intersync-delay" default: default-min-intersync-delay))
(if (and (not (args:get-arg "-sync-to-megatest.db")) ;; conditions under which we do not run the sync
(args:get-arg "-server"))
(let loop ()
(debug:print 0 *default-log-port* "INFO: syncer thread sleeping for server.minimum-intersync-delay seconds ["min-intersync-delay"]")
(thread-sleep! min-intersync-delay)
(if (common:simple-file-lock lockfile)
(begin
(if (not (configf:lookup *configdat* "server" "disable-db-snapshot"))
(common:snapshot-file mtdbfile subdir: ".db-snapshot"))
(delete-file* staging-file)
(let* ((start-time (current-milliseconds))
(res (system sync-cmd)))
(cond
((eq? 0 res)
(delete-file* (conc mtdbfile ".backup"))
(if (eq? 0 (file-size sync-log))
(delete-file sync-log))
(system (conc "/bin/mv " staging-file " " mtdbfile))
(debug:print 1 *default-log-port* "INFO: ["(common:human-time)"] pid="(current-process-id)" SYNC took "(/ (- (current-milliseconds) start-time) 1000)" sec")
#t)
(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)))))
(common:simple-file-release-lock lockfile)))
;; else
(debug:print 1 *default-log-port* "INFO: ["(common:human-time)"] pid="(current-process-id)" other SYNC in progress; not syncing.")
) ;; end if got lockfile
;; 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)
(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)
))))))
;; (let ((legacy-sync (common:run-sync?)))
;; (sync-stale-seconds (configf:lookup-number *configdat* "server" "sync-stale-seconds" default: 300))
;; (debug-mode (debug:debug-mode 1))
;; (last-time (current-seconds))
;; (no-sync-db (db:open-no-sync-db))
;; (sync-duration 0) ;; run time of the sync in milliseconds
;; ;;(this-wd-num (begin (mutex-lock! *wdnum*mutex) (let ((x *wdnum*)) (set! *wdnum* (add1 *wdnum*)) (mutex-unlock! *wdnum*mutex) x)))
;; )
;; (set! *no-sync-db* no-sync-db) ;; make the no sync db available to api calls
;; (debug:print-info 2 *default-log-port* "Periodic sync thread started.")
;; (debug:print-info 3 *default-log-port* "watchdog starting. legacy-sync is " legacy-sync" pid="(current-process-id) );; " this-wd-num="this-wd-num)
;; (if (and legacy-sync (not *time-to-exit*))
;; (let* (;;(dbstruct (db:setup))
;; (mtdb (dbr:dbstruct-mtdb dbstruct))
;; (mtpath (db:dbdat-get-path mtdb))
;; (tmp-area (common:get-db-tmp-area))
;; (lockfile (conc tmp-area "/megatest.db.lock"))
;; (start-file (conc tmp-area "/.start-sync"))
;; (end-file (conc tmp-area "/.end-sync")))
;; (debug:print-info 0 *default-log-port* "Server running, periodic sync started.")
;; (let loop ()
;; ;; sync for filesystem local db writes
;; ;;
;;
;; (mutex-lock! *db-multi-sync-mutex*)
;; (let* ((need-sync (>= *db-last-access* *db-last-sync*)) ;; no sync since last write
;; (sync-in-progress *db-sync-in-progress*)
;;
;; (min-intersync-delay (configf:lookup-number *configdat* "server" "minimum-intersync-delay" default: 5))
;; (should-sync (and (not *time-to-exit*)
;; (> (- (current-seconds) *db-last-sync*) min-intersync-delay))) ;; sync every five seconds minimum, deprecated logic, can probably be removed
;; (start-time (current-seconds))
;; (cpu-load-adj (alist-ref 'adj-proc-load (common:get-normalized-cpu-load #f)))
;; (mt-mod-time (file-modification-time mtpath))
;; (last-sync-start (if (common:file-exists? start-file)
;; (file-modification-time start-file)
;; 0))
;; (last-sync-end (if (common:file-exists? end-file)
;; (file-modification-time end-file)
;; 10))
;; (sync-period (+ 3 (* cpu-load-adj 30))) ;; as adjusted load increases increase the sync period
;; (recently-synced (and (< (- start-time mt-mod-time) sync-period) ;; not useful if sync didn't modify megatest.db!
;; (< mt-mod-time last-sync-start)))
;; (sync-done (<= last-sync-start last-sync-end))
;; (sync-stale (> start-time (+ last-sync-start sync-stale-seconds)))
;; (will-sync-pre (and (not *time-to-exit*) ;; do not start a sync if we are in the process of exiting
;; have-lock?
;; (or need-sync should-sync)
;; (or sync-done sync-stale)
;; (not sync-in-progress)
;; (not recently-synced)))
;; (will-sync (if will-sync-pre
;; ;; delay get lock until we decide to sync
;; #t ;; (server:have-sync-lock?)
;; #f)))
;; ;; if another server is syncing, postpone sync
;; (if (and will-sync-pre (not will-sync))
;; (set! *db-last-sync* start-time))
;; (debug:print-info 13 *default-log-port* "WD writable-watchdog top of loop. need-sync="need-sync" sync-in-progress=" sync-in-progress
;; " should-sync="should-sync" start-time="start-time" mt-mod-time="mt-mod-time" recently-synced="recently-synced" will-sync="will-sync
;; " sync-done=" sync-done " sync-period=" sync-period)
;; (if (and (> sync-period 5)
;; (common:low-noise-print 30 "sync-period"))
;; (debug:print-info 0 *default-log-port* "Increased sync period due to long sync times, sync took: " sync-period " seconds."))
;; ;; (if recently-synced (debug:print-info 0 *default-log-port* "Skipping sync due to recently-synced flag=" recently-synced))
;; ;; (debug:print-info 0 *default-log-port* "need-sync: " need-sync " sync-in-progress: " sync-in-progress " should-sync: " should-sync " will-sync: " will-sync)
;; (if will-sync (set! *db-sync-in-progress* #t))
;; (mutex-unlock! *db-multi-sync-mutex*)
;; (if will-sync
;; (let (;; (max-sync-duration (configf:lookup-number *configdat* "server" "max-sync-duration")) ;; KEEPING THIS AVAILABLE BUT SHOULD NOT USE, I'M PRETTY SURE IT DOES NOT WORK!
;; (sync-start (current-milliseconds)))
;; (with-output-to-file start-file (lambda ()(print (current-process-id))))
;;
;; ;; put lock here
;;
;; ;; (if (or (not max-sync-duration)
;; ;; (< sync-duration max-sync-duration)) ;; NOTE: db:sync-to-megatest.db keeps track of time of last sync and syncs incrementally
;; (let ((res (db:sync-to-megatest.db dbstruct no-sync-db: no-sync-db))) ;; did we sync any data? If so need to set the db touched flag to keep the server alive
;; (set! sync-duration (- (current-milliseconds) sync-start))
;; (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
;; ;; (server:release-sync-lock)
;; (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)
;; (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)))))))
|