38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
|
(import commonmod)
(declare (uses dbmod))
(import dbmod)
(declare (uses configfmod))
(import configfmod)
(include "common_records.scm")
(include "db_records.scm")
(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
;;======================================================================
;; ???
|
>
>
>
<
<
<
<
<
|
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
|
(import commonmod)
(declare (uses dbmod))
(import dbmod)
(declare (uses configfmod))
(import configfmod)
(declare (uses servermod))
(import servermod)
(include "common_records.scm")
(include "db_records.scm")
(define *server-loop-heart-beat* (current-seconds))
;;======================================================================
;; P K T S S T U F F
;;======================================================================
;; ???
|
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
122
123
124
125
126
127
128
129
130
131
|
(define (server:launch run-id transport-type)
(case transport-type
((http)(http-transport:launch))
;;((nmsg)(nmsg-transport:launch run-id))
;;((rpc) (rpc-transport:launch run-id))
(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)
(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)
(current-process-id)
(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))
|
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
|
77
78
79
80
81
82
83
84
85
86
87
88
89
90
|
(define (server:launch run-id transport-type)
(case transport-type
((http)(http-transport:launch))
;;((nmsg)(nmsg-transport:launch run-id))
;;((rpc) (rpc-transport:launch run-id))
(else (debug:print-error 0 *default-log-port* "unknown server type " transport-type))))
;; 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))
|
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
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
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
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
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
388
389
390
391
392
393
|
#;(common:wait-for-homehost-load load-limit (conc " delaying server start due to load on homehost. limit is " load-limit))
(system (conc "nbfake " cmdln))
(unsetenv "TARGETHOST_LOGF")
(if (get-environment-variable "TARGETHOST")(unsetenv "TARGETHOST"))
(thread-join! log-rotate)
(pop-directory)))
;; given a path to a server log return: host port startseconds
;; any changes to number of elements returned by this fuction will dirctly affect server:record->url,server:record->id,server:kill,server:get-num-alive which uses match let
(define (server:logf-get-start-info logf)
(let ((rx (regexp "^SERVER STARTED: (\\S+):(\\d+) AT ([\\d\\.]+) server-id: (\\S+)"))) ;; SERVER STARTED: host:port AT timesecs server id
(handle-exceptions
exn
(begin
(debug:print-info 0 *default-log-port* "Unable to get server info from " logf ", exn=" exn)
(list #f #f #f #f)) ;; no idea what went wrong, call it a bad server
(with-input-from-file
logf
(lambda ()
(let loop ((inl (read-line))
(lnum 0))
(if (not (eof-object? inl))
(let ((mlst (string-match rx inl)))
(if (not mlst)
(if (< lnum 500) ;; give up if more than 500 lines of server log read
(loop (read-line)(+ lnum 1))
(begin
(debug:print-info 0 *default-log-port* "Unable to get server info from first 500 lines of " logf )
(list #f #f #f #f)))
(let ((dat (cdr mlst)))
(list (car dat) ;; host
(string->number (cadr dat)) ;; port
(string->number (caddr dat))
(cadr (cddr dat))))))
(begin
(debug:print-info 0 *default-log-port* "Unable to get server info from " logf " at " (current-seconds))
(list #f #f #f #f)))))))))
;; get a list of servers with all relevant data
;; ( mod-time host port start-time pid )
;;
(define (server:get-list areapath #!key (limit #f))
(let ((fname-rx (regexp "^(|.*/)server-(\\d+)-(\\S+).log$"))
(day-seconds (* 24 60 60)))
;; if the directory exists continue to get the list
;; otherwise attempt to create the logs dir and then
;; continue
(if (if (directory-exists? (conc areapath "/logs"))
'()
(if (file-write-access? areapath)
(begin
(condition-case
(create-directory (conc areapath "/logs") #t)
(exn (i/o file)(debug:print 0 *default-log-port* "ERROR: Cannot create directory at " (conc areapath "/logs")))
(exn ()(debug:print 0 *default-log-port* "ERROR: Unknown error attemtping to get server list. exn=" exn)))
(directory-exists? (conc areapath "/logs")))
'()))
(let* ((server-logs (glob (conc areapath "/logs/server-*.log")))
(num-serv-logs (length server-logs)))
(if (null? server-logs)
'()
(let loop ((hed (car server-logs))
(tal (cdr server-logs))
(res '()))
(let* ((mod-time (handle-exceptions
exn
(begin
(debug:print 0 *default-log-port* "failed to get modification time on " hed ", exn=" exn)
(current-seconds)) ;; 0
(file-modification-time hed))) ;; default to *very* old so log gets ignored if deleted
(down-time (- (current-seconds) mod-time))
(serv-dat (if (or (< num-serv-logs 10)
(< down-time 900)) ;; day-seconds))
(server:logf-get-start-info hed)
'())) ;; don't waste time processing server files not touched in the 15 minutes if there are more than ten servers to look at
(serv-rec (cons mod-time serv-dat))
(fmatch (string-match fname-rx hed))
(pid (if fmatch (string->number (list-ref fmatch 2)) #f))
(new-res (if (null? serv-dat)
res
(cons (append serv-rec (list pid)) res)))) ;; any changes to number of elements in new-res will dirctly affect server:record->url,server:record->id,server:kill,server:get-num-alive which uses match let
(if (null? tal)
(if (and limit
(> (length new-res) limit))
new-res ;; (take new-res limit) <= need intelligent sorting before this will work
new-res)
(loop (car tal)(cdr tal) new-res)))))))))
(define (server:get-num-alive srvlst)
(let ((num-alive 0))
(for-each
(lambda (server)
(handle-exceptions
exn
(begin
(debug:print-info 0 *default-log-port* "Unable to get server start-time and/or mod-time from " server ", exn=" exn))
(match-let (((mod-time host port start-time server-id pid)
server))
(let* ((uptime (- (current-seconds) mod-time))
(runtime (if start-time
(- mod-time start-time)
0)))
(if (< uptime 5)(set! num-alive (+ num-alive 1)))))))
srvlst)
num-alive))
;; given a list of servers get a list of valid servers, i.e. at least
;; 10 seconds old, has started and is less than 1 hour old and is
;; active (i.e. mod-time < 10 seconds
;;
;; mod-time host port start-time pid
;;
;; sort by start-time descending. I.e. get the oldest first. Young servers will thus drop off
;; and servers should stick around for about two hours or so.
;;
(define (server:get-best srvlst)
(let* ((nums (server:get-num-servers))
(now (current-seconds))
(slst (sort
(filter (lambda (rec)
(if (and (list? rec)
(> (length rec) 2))
(let ((start-time (list-ref rec 3))
(mod-time (list-ref rec 0)))
;; (print "start-time: " start-time " mod-time: " mod-time)
(and start-time mod-time
(> (- now start-time) 0) ;; been running at least 0 seconds
(< (- now mod-time) 16) ;; still alive - file touched in last 16 seconds
(or (not (configf:lookup *configdat* "server" "runtime")) ;; skip if not set
(< (- now start-time)
(+ (- (string->number (configf:lookup *configdat* "server" "runtime"))
180)
(random 360)))) ;; under one hour running time +/- 180
))
#f))
srvlst)
(lambda (a b)
(< (list-ref a 3)
(list-ref b 3))))))
(if (> (length slst) nums)
(take slst nums)
slst)))
(define (server:get-first-best areapath)
(let ((srvrs (server:get-best (server:get-list areapath))))
(if (and srvrs
(not (null? srvrs)))
(car srvrs)
#f)))
(define (server:get-rand-best areapath)
(let ((srvrs (server:get-best (server:get-list areapath))))
(if (and (list? srvrs)
(not (null? srvrs)))
(let* ((len (length srvrs))
(idx (random len)))
(list-ref srvrs idx))
#f)))
(define (server:record->id servr)
(handle-exceptions
exn
(begin
(debug:print-info 0 *default-log-port* "Unable to get server id from " servr ", exn=" exn)
#f)
(match-let (((mod-time host port start-time server-id pid)
servr))
(if server-id
server-id
#f))))
(define (server:record->url servr)
(handle-exceptions
exn
(begin
(debug:print-info 0 *default-log-port* "Unable to get server url from " servr ", exn=" exn)
#f)
(match-let (((mod-time host port start-time server-id 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.
(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
;;
(define (server:wait-for-server-start-last-flag areapath)
(let* ((start-flag (conc areapath "/logs/server-start-last"))
;;; THIS INTERACTS WITH [server] timeout. Suggest using 0.1 or above for timeout (6 seconds)
(reftime (configf:lookup-number *configdat* "server" "idletime" default: 4))
(server-key (conc (get-host-name) "-" (current-process-id))))
(if (file-exists? start-flag)
(let* ((fmodtime (file-modification-time start-flag))
(delta (- (current-seconds) fmodtime))
(all-go (> delta reftime)))
(if (and all-go
(begin
(with-output-to-file start-flag
(lambda ()
(print server-key)))
(thread-sleep! 0.25)
(let ((res (with-input-from-file start-flag
(lambda ()
(read-line)))))
(equal? server-key res))))
#t ;; (system (conc "touch " start-flag)) ;; lazy but safe
(begin
(debug:print-info 0 *default-log-port* "Gating server start, last start: "
fmodtime ", delta: " delta ", reftime: " reftime ", all-go=" all-go)
(thread-sleep! reftime)
(server:wait-for-server-start-last-flag areapath)))))))
;; kind start up of servers, wait 40 seconds before allowing another server for a given
;; run-id to be launched
;;
(define (server:kind-run areapath)
;; look for $MT_RUN_AREA_HOME/logs/server-start-last
;; and wait for it to be at least 3 seconds old
|
|
|
|
|
|
|
|
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
|
|
<
|
|
<
<
<
<
|
<
>
|
<
<
<
<
<
|
<
<
<
<
|
|
<
<
<
<
|
<
|
<
<
|
<
<
<
<
<
<
<
<
|
<
<
<
|
<
<
<
<
<
<
<
<
<
<
<
<
<
|
|
<
<
<
<
<
<
|
<
|
<
<
<
|
>
|
>
|
>
|
<
|
|
<
|
<
<
|
|
<
<
<
<
<
|
<
|
<
<
<
<
<
|
<
<
<
<
<
<
<
<
<
<
<
<
|
<
<
|
<
<
|
|
|
<
<
<
<
<
<
|
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
182
183
184
185
186
187
188
189
190
191
|
#;(common:wait-for-homehost-load load-limit (conc " delaying server start due to load on homehost. limit is " load-limit))
(system (conc "nbfake " cmdln))
(unsetenv "TARGETHOST_LOGF")
(if (get-environment-variable "TARGETHOST")(unsetenv "TARGETHOST"))
(thread-join! log-rotate)
(pop-directory)))
(define (server:ping host-port-in server-id #!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))) ;; )
(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))
(port (cadr host-port))
(server-dat (http-transport:client-connect iface port server-id))
(login-res (rmt:login-no-auto-client-setup server-dat)))
(if (and (list? login-res)
(car login-res))
(begin
;; (print "LOGIN_OK")
(if do-exit (exit 0))
#t)
(begin
;; (print "LOGIN_FAILED")
(if do-exit (exit 1))
#f)))))))
;; kind start up of servers, wait 40 seconds before allowing another server for a given
;; run-id to be launched
;;
(define (server:kind-run areapath)
;; look for $MT_RUN_AREA_HOME/logs/server-start-last
;; and wait for it to be at least 3 seconds old
|
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
|
(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.
;;
(define (server:check-if-running areapath) ;; #!key (numservers "2"))
(let* ((ns (server:get-num-servers))
(servers (server:get-best (server:get-list areapath))))
(if (or (and servers
|
<
<
<
<
<
<
|
223
224
225
226
227
228
229
230
231
232
233
234
235
236
|
(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)))))))
;; no longer care if multiple servers are started by accident. older servers will drop off in time.
;;
(define (server:check-if-running areapath) ;; #!key (numservers "2"))
(let* ((ns (server:get-num-servers))
(servers (server:get-best (server:get-list areapath))))
(if (or (and servers
|
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
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
|
((http)(server:ping server-url server-id))
;; ((nmsg)(nmsg-transport:ping (tasks:hostinfo-get-interface server)
)))
(if res
server-url
#f)))
(define (server:kill servr)
(handle-exceptions
exn
(begin
(debug:print-info 0 *default-log-port* "Unable to get host and/or port from " servr ", exn=" exn)
#f)
(match-let (((mod-time hostname port start-time server-id pid)
servr))
(tasks:kill-server hostname pid))))
;; called in megatest.scm, host-port is string hostname:port
;;
;; 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 server-id #!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))) ;; )
(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))
(port (cadr host-port))
(server-dat (http-transport:client-connect iface port server-id))
(login-res (rmt:login-no-auto-client-setup server-dat)))
(if (and (list? login-res)
(car login-res))
(begin
;; (print "LOGIN_OK")
(if do-exit (exit 0))
#t)
(begin
;; (print "LOGIN_FAILED")
(if do-exit (exit 1))
#f)))))))
;; run ping in separate process, safest way in some cases
;;
(define (server:ping-server ifaceport)
(with-input-from-pipe
(conc (common:get-megatest-exe) " -ping " ifaceport)
(lambda ()
(let loop ((inl (read-line))
(res "NOREPLY"))
(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))
60)))
(define (server:get-best-guess-address hostname)
(let ((res #f))
(for-each
(lambda (adr)
(if (not (eq? (u8vector-ref adr 0) 127))
(set! res adr)))
;; 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))
(tmp-db (conc tmp-area "/megatest.db"))
(staging-file (conc *toppath* "/.megatest.db"))
(mtdbfile (conc *toppath* "/megatest.db"))
(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))
(default-duty-cycle 0.1)
(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)
(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
" , off-time="off-time" seconds ]")
(thread-sleep! (max off-time min-intersync-delay)))
(else
(debug:print 0 *default-log-port* "INFO: syncer thread NOT sleeping ; maybe time-to-exit...")))
(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))
(dbbackupfile (conc mtdbfile ".backup"))
(res2
(cond
((eq? 0 res )
(handle-exceptions
exn
#f
(if (file-exists? dbbackupfile)
(delete-file* dbbackupfile)
)
(if (eq? 0 (file-size sync-log))
(delete-file* sync-log))
(system (conc "/bin/mv " staging-file " " mtdbfile))
(set! last-sync-seconds (/ (- (current-milliseconds) start-time) 1000))
(set! off-time (calculate-off-time
last-sync-seconds
(cond
((and (number? duty-cycle) (> duty-cycle 0) (< duty-cycle 1))
duty-cycle)
(else
(debug:print 0 *default-log-port* "WARNING: ["(common:human-time)"] server.sync-duty-cycle is invalid. Should be a number between 0 and 1, but "duty-cycle" was specified. Using default value: "default-duty-cycle)
default-duty-cycle))))
(debug:print 1 *default-log-port* "INFO: ["(common:human-time)"] pid="(current-process-id)" SYNC took "last-sync-seconds" sec")
(debug:print 1 *default-log-port* "INFO: ["(common:human-time)"] pid="(current-process-id)" SYNC took "last-sync-seconds" sec ; with duty-cycle of "duty-cycle" off time is now "off-time)
'sync-completed))
(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
(let* ((do-a-sync (server:get-bruteforce-syncer dbstruct))
(final-sync (server:get-bruteforce-syncer dbstruct fork-to-background: #t persist-until-sync: #t)))
(when (and (not (args:get-arg "-sync-to-megatest.db")) ;; conditions under which we do not run the sync
(args:get-arg "-server"))
|
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
|
|
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
|
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
|
<
<
<
|
<
<
<
<
<
<
<
<
<
|
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
|
<
<
<
<
<
<
|
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
|
((http)(server:ping server-url server-id))
;; ((nmsg)(nmsg-transport:ping (tasks:hostinfo-get-interface server)
)))
(if res
server-url
#f)))
(define server:try-running server:run) ;; there is no more per-run servers ;; REMOVE ME. BUG.
(define (server:writable-watchdog-bruteforce dbstruct)
(thread-sleep! 1) ;; delay for startup
(let* ((do-a-sync (server:get-bruteforce-syncer dbstruct))
(final-sync (server:get-bruteforce-syncer dbstruct fork-to-background: #t persist-until-sync: #t)))
(when (and (not (args:get-arg "-sync-to-megatest.db")) ;; conditions under which we do not run the sync
(args:get-arg "-server"))
|