45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
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
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
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
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
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
|
;;======================================================================
;; P K T S S T U F F
;;======================================================================
;; ???
;;======================================================================
;; S E R V E R
;;======================================================================
;; Call this to start the actual server
;;
;; all routes though here end in exit ...
;;
;; start_server
;;
(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
;;======================================================================
;; 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)))
#;(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*)))
(define server:try-running server:run) ;; there is no more per-run servers ;; REMOVE ME. BUG.
(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))))
;; 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)))
;; (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"))
(let loop ()
(do-a-sync)
(if (not (bdat-time-to-exit *bdat*)) (loop))) ;; keep going unless time to exit
;; time to exit, close the no-sync db here
(final-sync)
(if (common:low-noise-print 30)
(debug:print-info 0 *default-log-port* "Exiting watchdog timer, *time-to-exit* = "(bdat-time-to-exit *bdat*)" pid="(current-process-id)
)))))
(define (server:writable-watchdog-deltasync dbstruct)
(thread-sleep! 0.054) ;; delay for startup
(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))
(stmt-cache (dbr:dbstruct-stmt-cache dbstruct))
(sync-duration 0) ;; run time of the sync in milliseconds
)
(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 (bdat-time-to-exit *bdat*)))
(let* (;;(dbstruct (db:setup))
(mtdb (dbr:dbstruct-mtdb dbstruct))
(mtpath (db:dbdat-get-path mtdb))
(tmp-area (common:get-db-tmp-area))
(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 (bdat-time-to-exit *bdat*))
(> (- (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 (and (not (bdat-time-to-exit *bdat*)) ;; do not start a sync if we are in the process of exiting
(or need-sync should-sync)
(or sync-done sync-stale)
(not sync-in-progress)
(not recently-synced))))
(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
(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 (bdat-time-to-exit *bdat*))
(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*="(bdat-time-to-exit *bdat*))
(if (and (not (bdat-time-to-exit *bdat*))
(< count 6)) ;; was 11, changing to 4.
(begin
(thread-sleep! 1)
(delay-loop (+ count 1))))
(if (not (bdat-time-to-exit *bdat*)) (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* = " (bdat-time-to-exit *bdat*)" pid="(current-process-id) ))))))) ;;" this-wd-num="this-wd-num)))))))
|
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
|
45
46
47
48
49
50
51
|
;;======================================================================
;; P K T S S T U F F
;;======================================================================
;; ???
|