Megatest

Diff
Login

Differences From Artifact [af79560081]:

To Artifact [6f2c907335]:


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
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
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
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
(define (tasks:hostinfo-get-interface   vec)    (vector-ref  vec 1))
(define (tasks:hostinfo-get-port        vec)    (vector-ref  vec 2))
(define (tasks:hostinfo-get-pubport     vec)    (vector-ref  vec 3))
(define (tasks:hostinfo-get-transport   vec)    (vector-ref  vec 4))
(define (tasks:hostinfo-get-pid         vec)    (vector-ref  vec 5))
(define (tasks:hostinfo-get-hostname    vec)    (vector-ref  vec 6))

(define (tasks:server-lock-slot mdb run-id)
  (tasks:server-clean-out-old-records-for-run-id mdb run-id " tasks:server-lock-slot")
  (if (< (tasks:num-in-available-state mdb run-id) 4)
      (begin 
	(tasks:server-set-available mdb run-id)
	(thread-sleep! (/ (random 1500) 1000)) ;; (thread-sleep! 2) ;; Try removing this. It may not be needed.
	(tasks:server-am-i-the-server? mdb run-id))
      #f))
	
;; register that this server may come online (first to register goes though with the process)
(define (tasks:server-set-available mdb run-id)
  (sqlite3:execute 
   mdb 
   "INSERT INTO servers (pid,hostname,port,pubport,start_time,      priority,state,mt_version,heartbeat,   interface,transport,run_id)
                   VALUES(?, ?,       ?,   ?, strftime('%s','now'), ?,       ?,    ?,-1,?,        ?,        ?);"
   (current-process-id)          ;; pid
   (get-host-name)               ;; hostname
   -1                            ;; port
   -1                            ;; pubport
   (random 1000)                 ;; priority (used a tiebreaker on get-available)
   "available"                   ;; state
   (common:version-signature)    ;; mt_version
   -1                            ;; interface
   ;; (conc (server:get-transport)) ;; transport
   (conc *transport-type*)    ;; transport
   run-id
   ))

(define (tasks:num-in-available-state mdb run-id)
  (let ((res 0))
    (sqlite3:for-each-row
     (lambda (num-in-queue)
       (set! res num-in-queue))
     mdb
     "SELECT count(id) FROM servers WHERE run_id=? AND state = 'available' AND (strftime('%s','now') - start_time) < 30 ;"
     run-id)
    res))

(define (tasks:num-servers-non-zero-running mdb)
  (let ((res 0))
    (sqlite3:for-each-row
     (lambda (num-running)
       (set! res num-running))
     mdb
     "SELECT count(id) FROM servers WHERE run_id != 0 AND state = 'running';")
    res))

(define (tasks:server-clean-out-old-records-for-run-id mdb run-id tag)
  (sqlite3:execute mdb "UPDATE servers SET state=?,heartbeat=strftime('%s','now') WHERE state in ('available','dbprep','shutting-down') AND (strftime('%s','now') - start_time) > 50 AND run_id=?;"
		   (conc "defunct" tag) run-id))

(define (tasks:server-force-clean-running-records-for-run-id mdb run-id tag)
  (sqlite3:execute mdb "UPDATE servers SET state=?,heartbeat=strftime('%s','now') WHERE state = 'running' AND run_id=?;"
		   (conc "defunct" tag) run-id))

(define (tasks:server-force-clean-run-record mdb run-id iface port tag)
  (sqlite3:execute mdb "UPDATE servers SET state=?,heartbeat=strftime('%s','now') WHERE state = 'running' AND run_id=? AND interface=? AND port=?;"
		   (conc "defunct" tag) run-id iface port))


;; BB> adding missing func for --list-servers
(define (tasks:server-deregister mdb hostname #!key (pullport #f) (pid #f) (action #f)) ;;pullport pid: pid action: 'delete))
  (if (eq? action 'delete)
      (sqlite3:execute mdb "DELETE FROM servers WHERE pid=? AND port=? AND hostname=?;" pid pullport hostname)
      (sqlite3:execute mdb "UPDATE servers SET state='defunct', heartbeat=strftime('%s','now') WHERE hostname=? AND pid=?;"
                       hostname pid)))

(define (tasks:server-delete-records-for-this-pid mdb tag)
  (sqlite3:execute mdb "UPDATE servers SET state=?,heartbeat=strftime('%s','now') WHERE hostname=? AND pid=?;"
		   (conc "defunct" tag) (get-host-name) (current-process-id)))

(define (tasks:server-delete-record mdb server-id tag) 
  (sqlite3:execute mdb "UPDATE servers SET state=?,heartbeat=strftime('%s','now') WHERE id=?;"
		   (conc "defunct" tag) server-id)
  ;; use this opportuntity to clean out records over one month old or over 10 minutes old with port = -1 (i.e. a never used placeholder)
  (sqlite3:execute mdb "DELETE FROM servers WHERE state not in ('running','shutting-down','dbprep') AND (strftime('%s','now') - start_time) > 2628000;")
  (sqlite3:execute mdb "DELETE FROM servers WHERE state like 'defunct%' AND port=-1 AND (strftime('%s','now') - start_time) > 600;")
  )

(define (tasks:server-set-state! mdb server-id state)
  (sqlite3:execute mdb "UPDATE servers SET state=?,heartbeat=strftime('%s','now') WHERE id=?;" state server-id))

(define (tasks:server-set-interface-port mdb server-id interface port)
  (sqlite3:execute mdb "UPDATE servers SET interface=?,port=?,heartbeat=strftime('%s','now') WHERE id=?;" interface port server-id))

;; Get random port not used in long time
;;
(define (tasks:server-get-next-port mdb)
  (let* ((lownum        30000)
	(highnum        64000)
	(used-ports     '())
	(get-rand-port  (lambda ()
			  (+ lownum (random (- highnum lownum)))))
	(port-param     (if (and (args:get-arg "-port")
				 (string->number (args:get-arg "-port")))
			    (string->number (args:get-arg "-port"))
			    #f))
	;; (config-port    (if (and (config-lookup  *configdat* "server" "port")
	;; 			 (string->number (config-lookup  *configdat* "server" "port")))
	;; 		    (string->number (config-lookup  *configdat* "server" "port"))
	;; 		    #f))
	)
    (sqlite3:for-each-row
     (lambda (port)
       (set! used-ports (cons port used-ports)))
     mdb
     "SELECT port FROM servers;")
    (cond
     ((and port-param res)   (if (> res port-param) res port-param))
     (port-param             port-param)
     ;; ((and config-port res)  (if (> res config-port) res config-port))
     ;; (config-port            config-port)
     (else
      (let loop ((port     (get-rand-port))
		 (remtries 100))
	(if (member port used-ports)
	    (if (> remtries 0)
		(loop (get-rand-port)(- remtries 1))
		(get-rand-port))
	    port))))))

(define (tasks:server-am-i-the-server? mdb run-id)
  (let* ((all    (tasks:server-get-servers-vying-for-run-id mdb run-id))
	 (first  (if (null? all)
		     #f;; (begin (debug:print-error 0 *default-log-port* "no servers listed, should be at least one by now.") 
		       ;;      (sqlite3:finalize! mdb)
		       ;;      (exit 1))
		     (car (db:get-rows all)))))
    (if first
	(let* ((header   (db:get-header all))
	       (id       (db:get-value-by-header first header "id"))
	       (hostname (db:get-value-by-header first header "hostname"))
	       (pid      (db:get-value-by-header first header "pid"))
	       (priority (db:get-value-by-header first header "priority")))
	  ;; (debug:print 0 *default-log-port* "INFO: am-i-the-server got record " first)
	  ;; for now a basic check. add tiebreaking by priority later
	  (if (and (equal? hostname (get-host-name))
		   (equal? pid      (current-process-id)))
	      id
	      #f))
	#f)))
	     
;; Use: (db:get-value-by-header (car (db:get-rows dat)) (db:get-header dat) "fieldname")
;;  to extract info from the structure returned
;;
(define (tasks:server-get-servers-vying-for-run-id mdb run-id)
   (let* ((header (list "id" "hostname" "pid" "interface" "port" "pubport" "state" "run_id" "priority" "start_time"))
	  (selstr (string-intersperse header ","))
	  (res    '()))
    (sqlite3:for-each-row
     (lambda (a . b)
       (set! res (cons (apply vector a b) res)))
     mdb
     (conc "SELECT " selstr " FROM servers WHERE state in ('available','running','dbprep') ORDER BY start_time DESC;")
     )
    (vector header res)))

(define (tasks:get-server mdb run-id #!key (retries 10))
  (let ((res  #f)
	(best #f))
    (handle-exceptions
     exn
     (begin
       (print-call-chain (current-error-port))
       (debug:print 0 *default-log-port* "WARNING: tasks:get-server db access error.")
       (debug:print 0 *default-log-port* " message: " ((condition-property-accessor 'exn 'message) exn))
       (debug:print 0 *default-log-port* " for run " run-id)
       (print-call-chain (current-error-port))
       (if (> retries 0)
	   (begin
	     (debug:print 0 *default-log-port* " trying call to tasks:get-server again in 10 seconds")
	     (thread-sleep! 10)
	     (tasks:get-server mdb run-id retries: (- retries 0)))
	   (debug:print 0 *default-log-port* "10 tries of tasks:get-server all crashed and burned. Giving up and returning \"no server found\"")))
     (sqlite3:for-each-row
      (lambda (id interface port pubport transport pid hostname)
	(set! res (vector id interface port pubport transport pid hostname)))
      mdb
      ;; removed:
      ;; strftime('%s','now')-heartbeat < 10 AND mt_version = ?
      "SELECT id,interface,port,pubport,transport,pid,hostname FROM servers
          WHERE run_id=? AND state='running'
          ORDER BY start_time DESC LIMIT 1;" run-id) ;; (common:version-signature) run-id)
     res)))

(define (tasks:server-running-or-starting? mdb run-id)
  (let ((res #f))
    (sqlite3:for-each-row
     (lambda (id)
       (set! res id))
     mdb ;; NEEDS dbprep ADDED
     "SELECT id FROM servers WHERE run_id=? AND (state = 'running' OR (state = 'dbprep' AND  (strftime('%s','now') - start_time) < 60));" run-id)
    res))

(define (tasks:server-running? mdb run-id)
  (let ((res #f))
    (sqlite3:for-each-row
     (lambda (id)
       (set! res id))
     mdb ;; NEEDS dbprep ADDED
     "SELECT id FROM servers WHERE run_id=? AND state = 'running';" run-id)
    res))

(define (tasks:need-server run-id)
  (equal? (configf:lookup *configdat* "server" "required") "yes"))

;; 	(maxqry (cdr (rmt:get-max-query-average run-id)))
;; 	(threshold   (string->number (or (configf:lookup *configdat* "server" "server-query-threshold") "10"))))
;;     (cond
;;      (forced 
;;       (if (common:low-noise-print 60 run-id "server required is set")
;; 	  (debug:print-info 0 *default-log-port* "Server required is set, starting server for run-id " run-id "."))
;;       #t)
;;      ((> maxqry threshold)
;;       (if (common:low-noise-print 60 run-id "Max query time execeeded")
;; 	  (debug:print-info 0 *default-log-port* "Max avg query time of " maxqry "ms exceeds limit of " threshold "ms, server needed for run-id " run-id "."))
;;       #t)
;;      (else
;;       #f))))

;; try to start a server and wait for it to be available
;;
(define (tasks:start-and-wait-for-server tdbdat run-id delay-max-tries)
  ;; ensure a server is running for this run
  (let loop ((server-dat (tasks:get-server (db:delay-if-busy tdbdat) run-id))
	     (delay-time 0))
      (if (and (not server-dat)
	       (< delay-time delay-max-tries))
	  (begin
	    (if (common:low-noise-print 60 "tasks:start-and-wait-for-server" run-id)
		(debug:print 0 *default-log-port* "Try starting server for run-id " run-id))
	    (thread-sleep! (/ (random 2000) 1000))
	    (server:kind-run *toppath*)
	    (thread-sleep! (min delay-time 1))
            (if (not (or (server:start-attempted? *toppath*)
                         (server:read-dotserver *toppath*))) ;; no point in trying
                (loop (tasks:get-server (db:delay-if-busy tdbdat) run-id)(+ delay-time 1))
                #f))
          #f)))

(define (tasks:get-all-servers mdb)
  (let ((res '()))
    (sqlite3:for-each-row
     (lambda (id pid hostname interface port pubport start-time priority state mt-version last-update transport run-id)
       ;;                       0  1     2         3      4     5          6        7     8          9          10        11     12
       (set! res (cons (vector id pid hostname interface port pubport start-time priority state mt-version last-update transport run-id) res)))
     mdb
     "SELECT id,pid,hostname,interface,port,pubport,start_time,priority,state,mt_version,strftime('%s','now')-heartbeat AS last_update,transport,run_id 
        FROM servers WHERE state NOT LIKE 'defunct%' ORDER BY start_time DESC;")
    res))

(define (tasks:get-server-by-id mdb id)
  (let ((res #f))
    (sqlite3:for-each-row
     (lambda (id pid hostname interface port pubport start-time priority state mt-version last-update transport run-id)
       ;;                       0  1     2         3      4     5          6        7     8          9          10        11     12
       (set! res (vector id pid hostname interface port pubport start-time priority state mt-version last-update transport run-id)))
     mdb
     "SELECT id,pid,hostname,interface,port,pubport,start_time,priority,state,mt_version,strftime('%s','now')-heartbeat AS last_update,transport,run_id 
        FROM servers WHERE id=?;"
     id)
    res))

(define (tasks:get-server-records mdb run-id)
  (let ((res '()))
    (sqlite3:for-each-row
     (lambda (id pid hostname interface port pubport start-time priority state mt-version last-update transport run-id)
       ;;                       0  1     2         3      4     5          6        7     8          9          10        11     12
       (set! res (cons (vector id pid hostname interface port pubport start-time priority state mt-version last-update transport run-id) res)))
     mdb
     "SELECT id,pid,hostname,interface,port,pubport,start_time,priority,state,mt_version,strftime('%s','now')-heartbeat AS last_update,transport,run_id 
        FROM servers WHERE run_id=? AND state NOT LIKE 'defunct%' ORDER BY start_time DESC;"
     run-id)
    (reverse res)))

;; no elegance here ...
;;
(define (tasks:kill-server hostname pid #!key (kill-switch ""))
  (server:remove-dotserver-file *toppath* ".*")
  (debug:print-info 0 *default-log-port* "Attempting to kill server process " pid " on host " hostname)
  (setenv "TARGETHOST" hostname)
  (setenv "TARGETHOST_LOGF" "server-kills.log")
  (system (conc "nbfake kill "kill-switch" "pid))

  (unsetenv "TARGETHOST_LOGF")
  (unsetenv "TARGETHOST"))
 
;; look up a server by run-id and send it a kill, also delete the record for that server
;;
(define (tasks:kill-server-run-id run-id #!key (tag "default"))
  (let* ((tdbdat  (tasks:open-db))
	 (sdat    (tasks:get-server (db:delay-if-busy tdbdat) run-id)))
    (if sdat
	(let ((hostname (vector-ref sdat 6))
	      (pid      (vector-ref sdat 5))
	      (server-id (vector-ref sdat 0)))
	  (tasks:server-set-state! (db:delay-if-busy tdbdat) server-id "killed")
	  (debug:print-info 0 *default-log-port* "Killing server " server-id " for run-id " run-id " on host " hostname " with pid " pid)
          (server:remove-dotserver-file *toppath* ".*")
	  (tasks:kill-server hostname pid)
	  (tasks:server-delete-record (db:delay-if-busy tdbdat) server-id tag) )
	(debug:print-info 0 *default-log-port* "No server found for run-id " run-id ", nothing to kill"))
    ;; (sqlite3:finalize! tdb)
    ))
    
;;======================================================================
;; M O N I T O R S
;;======================================================================

(define (tasks:remove-monitor-record mdb)
  (sqlite3:execute mdb "DELETE FROM monitors WHERE pid=? AND hostname=?;"
		   (current-process-id)







<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<



<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<



<








<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<







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
(define (tasks:hostinfo-get-interface   vec)    (vector-ref  vec 1))
(define (tasks:hostinfo-get-port        vec)    (vector-ref  vec 2))
(define (tasks:hostinfo-get-pubport     vec)    (vector-ref  vec 3))
(define (tasks:hostinfo-get-transport   vec)    (vector-ref  vec 4))
(define (tasks:hostinfo-get-pid         vec)    (vector-ref  vec 5))
(define (tasks:hostinfo-get-hostname    vec)    (vector-ref  vec 6))












































































































































































































(define (tasks:need-server run-id)
  (equal? (configf:lookup *configdat* "server" "required") "yes"))






































































;; no elegance here ...
;;
(define (tasks:kill-server hostname pid #!key (kill-switch ""))

  (debug:print-info 0 *default-log-port* "Attempting to kill server process " pid " on host " hostname)
  (setenv "TARGETHOST" hostname)
  (setenv "TARGETHOST_LOGF" "server-kills.log")
  (system (conc "nbfake kill "kill-switch" "pid))

  (unsetenv "TARGETHOST_LOGF")
  (unsetenv "TARGETHOST"))
 


















;;======================================================================
;; M O N I T O R S
;;======================================================================

(define (tasks:remove-monitor-record mdb)
  (sqlite3:execute mdb "DELETE FROM monitors WHERE pid=? AND hostname=?;"
		   (current-process-id)