| ︙ | | |
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
|
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
|
-
+
-
+
+
-
+
|
'())))
(define (launch:execute encoded-cmd)
(let* ((cmdinfo (read (open-input-string (base64:base64-decode encoded-cmd)))))
(setenv "MT_CMDINFO" encoded-cmd)
(if (list? cmdinfo) ;; ((testpath /tmp/mrwellan/jazzmind/src/example_run/tests/sqlitespeed)
;; (test-name sqlitespeed) (runscript runscript.rb) (db-host localhost) (run-id 1))
(let* ((testpath (assoc/default 'testpath cmdinfo)) ;; How is testpath different from work-area ??
(let* ((testpath (assoc/default 'testpath cmdinfo)) ;; testpath is the test spec area
(top-path (assoc/default 'toppath cmdinfo))
(work-area (assoc/default 'work-area cmdinfo))
(work-area (assoc/default 'work-area cmdinfo)) ;; work-area is the test run area
(test-name (assoc/default 'test-name cmdinfo))
(runscript (assoc/default 'runscript cmdinfo))
(ezsteps (assoc/default 'ezsteps cmdinfo))
;; (runremote (assoc/default 'runremote cmdinfo))
(transport (assoc/default 'transport cmdinfo))
(serverinf (assoc/default 'serverinf cmdinfo))
(port (assoc/default 'port cmdinfo))
(run-id (assoc/default 'run-id cmdinfo))
(test-id (assoc/default 'test-id cmdinfo))
(target (assoc/default 'target cmdinfo))
(itemdat (assoc/default 'itemdat cmdinfo))
(env-ovrd (assoc/default 'env-ovrd cmdinfo))
(set-vars (assoc/default 'set-vars cmdinfo)) ;; pre-overrides from -setvar
(runname (assoc/default 'runname cmdinfo))
(megatest (assoc/default 'megatest cmdinfo))
(runtlim (assoc/default 'runtlim cmdinfo))
(mt-bindir-path (assoc/default 'mt-bindir-path cmdinfo))
(keys #f)
(keyvals #f)
(fullrunscript (if (not runscript)
#f
(if (substring-index "/" runscript)
runscript ;; use unadultered if contains slashes
(let ((fulln (conc testpath "/" runscript)))
(if (and (file-exists? fulln)
(file-execute-access? fulln))
fulln
runscript))))) ;; assume it is on the path
(rollup-status 0))
(change-directory top-path)
(debug:print 2 "Exectuing " test-name " (id: " test-id ") on " (get-host-name))
;; Setup the *runremote* global var
(if *runremote* (debug:print 2 "ERROR: I'm not expecting *runremote* to be set at this time"))
;; (set! *runremote* runremote)
(set! *transport-type* (string->symbol transport))
(set! keys (cdb:remote-run db:get-keys #f))
(set! keyvals (if run-id (cdb:remote-run db:get-key-vals #f run-id) #f))
(set! keyvals (keys:target->keyval keys target))
;; apply pre-overrides before other variables. The pre-override vars must not
;; clobbers things from the official sources such as megatest.config and runconfigs.config
(if (string? set-vars)
(let ((varpairs (string-split set-vars ",")))
(debug:print 4 "varpairs: " varpairs)
(map (lambda (varpair)
(let ((varval (string-split varpair "=")))
|
| ︙ | | |
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
|
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
|
-
+
-
+
|
;; Can setup as client for server mode now
;; (client:setup)
(change-directory *toppath*)
(set-megatest-env-vars run-id) ;; these may be needed by the launching process
(change-directory work-area)
(open-run-close set-run-config-vars #f run-id keys keyvals)
(set-run-config-vars run-id keyvals target) ;; (db:get-target db run-id))
;; environment overrides are done *before* the remaining critical envars.
(alist->env-vars env-ovrd)
(set-megatest-env-vars run-id)
(set-item-env-vars itemdat)
(save-environment-as-files "megatest")
;; open-run-close not needed for test-set-meta-info
(test-set-meta-info #f test-id run-id test-name itemdat 0)
(tests:set-meta-info #f test-id run-id test-name itemdat 0 work-area)
(tests:test-set-status! test-id "REMOTEHOSTSTART" "n/a" (args:get-arg "-m") #f)
(if (args:get-arg "-xterm")
(set! fullrunscript "xterm")
(if (and fullrunscript (not (file-execute-access? fullrunscript)))
(system (conc "chmod ug+x " fullrunscript))))
;; We are about to actually kick off the test
;; so this is a good place to remove the records for
|
| ︙ | | |
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
|
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
|
-
+
-
+
|
;; (set! script (conc script "source " prev-env))))
;; call the command using mt_ezstep
(set! script (conc "mt_ezstep " stepname " " (if prevstep prevstep "-") " " stepcmd))
(debug:print 4 "script: " script)
;; DO NOT remote
(db:teststep-set-status! #f test-id stepname "start" "-" #f #f)
(db:teststep-set-status! #f test-id stepname "start" "-" #f #f work-area: work-area)
;; now launch
(let ((pid (process-run script)))
(let processloop ((i 0))
(let-values (((pid-val exit-status exit-code)(process-wait pid #t)))
(mutex-lock! m)
(vector-set! exit-info 0 pid)
(vector-set! exit-info 1 exit-status)
(vector-set! exit-info 2 exit-code)
(mutex-unlock! m)
(if (eq? pid-val 0)
(begin
(thread-sleep! 2)
(processloop (+ i 1))))
))
(let ((exinfo (vector-ref exit-info 2))
(logfna (if logpro-used (conc stepname ".html") "")))
;; testing if procedures called in a remote call cause problems (ans: no or so I suspect)
(db:teststep-set-status! #f test-id stepname "end" exinfo #f logfna))
(db:teststep-set-status! #f test-id stepname "end" exinfo #f logfna work-area: work-area))
(if logpro-used
(cdb:test-set-log! *runremote* test-id (conc stepname ".html")))
;; set the test final status
(let* ((this-step-status (cond
((and (eq? (vector-ref exit-info 2) 2) logpro-used) 'warn)
((eq? (vector-ref exit-info 2) 0) 'pass)
(else 'fail)))
|
| ︙ | | |
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
|
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
|
-
+
+
+
+
+
+
+
+
-
+
+
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
|
(round
(-
(current-seconds)
start-seconds)))))
(kill-tries 0))
(let loop ((minutes (calc-minutes)))
(begin
(set! kill-job? (test-get-kill-request test-id)) ;; run-id test-name itemdat))
(set! kill-job? (or (test-get-kill-request test-id) ;; run-id test-name itemdat))
(and runtlim (let* ((run-seconds (- (current-seconds) start-seconds))
(time-exceeded (> run-seconds runtlim)))
(if time-exceeded
(begin
(debug:print-info 0 "KILLING TEST DUE TO TIME LIMIT EXCEEDED! Runtime=" run-seconds " seconds, limit=" runtlim)
#t)
#f)))))
;; open-run-close not needed for test-set-meta-info
(test-set-meta-info #f test-id run-id test-name itemdat minutes)
(tests:set-meta-info #f test-id run-id test-name itemdat minutes work-area)
(if kill-job?
(begin
(mutex-lock! m)
(let* ((pid (vector-ref exit-info 0)))
(if (number? pid)
(process-signal pid signal/kill)
(begin
(debug:print 0 "WARNING: Request received to kill job (attempt # " kill-tries ")")
(let ((processes (cmd-run->list (conc "pgrep -l -P " pid))))
(for-each
(lambda (p)
(let* ((parts (string-split p))
(p-id (if (> (length parts) 0)
(string->number (car parts))
#f)))
(if p-id
(begin
(debug:print 0 "Killing " (cadr parts) "; kill -9 " p-id)
(system (conc "kill -9 " p-id))))))
(car processes))
(system (conc "kill -9 -" pid))))
;; (begin
;; (debug:print 0 "WARNING: Request received to kill job (attempt # " kill-tries ")")
;; (let ((processes (cmd-run->list (conc "pgrep -l -P " pid))))
;; (for-each
;; (lambda (p)
;; (let* ((parts (string-split p))
;; (p-id (if (> (length parts) 0)
;; (string->number (car parts))
;; #f)))
;; (if p-id
;; (begin
;; (debug:print 0 "Killing " (cadr parts) "; kill -9 " p-id)
;; (system (conc "kill -9 " p-id))))))
;; (car processes))
;; (system (conc "kill -9 -" pid))))
(begin
(debug:print 0 "WARNING: Request received to kill job but problem with process, attempting to kill manager process")
(tests:test-set-status! test-id "KILLED" "FAIL"
(args:get-arg "-m") #f)
(sqlite3:finalize! tdb)
(exit 1))))
(set! kill-tries (+ 1 kill-tries))
|
| ︙ | | |
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
|
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
|
-
-
+
+
-
-
-
-
+
+
+
+
-
-
+
+
+
+
+
+
+
+
+
+
+
-
+
|
(thread-start! th1)
(thread-start! th2)
(thread-join! th2)
(mutex-lock! m)
(let* ((item-path (item-list->path itemdat))
(testinfo (cdb:get-test-info-by-id *runremote* test-id))) ;; )) ;; run-id test-name item-path)))
;; Am I completed?
(if (not (equal? (db:test-get-state testinfo) "COMPLETED"))
(begin
(if (equal? (db:test-get-state testinfo) "RUNNING") ;; (not (equal? (db:test-get-state testinfo) "COMPLETED"))
(let ((new-state (if kill-job? "KILLED" "COMPLETED") ;; (if (eq? (vector-ref exit-info 2) 0) ;; exited with "good" status
(debug:print 2 "Test NOT logged as COMPLETED, (state=" (db:test-get-state testinfo) "), updating result, rollup-status is " rollup-status)
(tests:test-set-status! test-id
(if kill-job? "KILLED" "COMPLETED")
(cond
(else "FAIL"))
(args:get-arg "-m") #f)))
(else "FAIL")))) ;; (db:test-get-status testinfo)))
(debug:print-info 2 "Test NOT logged as COMPLETED, (state=" (db:test-get-state testinfo) "), updating result, rollup-status is " rollup-status)
(tests:test-set-status! test-id
new-state
new-status
(args:get-arg "-m") #f)
;; need to update the top test record if PASS or FAIL and this is a subtest
(if (not (equal? item-path ""))
(cdb:roll-up-pass-fail-counts *runremote* run-id test-name item-path new-status))
))
;; for automated creation of the rollup html file this is a good place...
(if (not (equal? item-path ""))
(open-run-close tests:summarize-items #f run-id test-name #f)) ;; don't force - just update if no
(tests:summarize-items #f run-id test-name #f)) ;; don't force - just update if no
)
(mutex-unlock! m)
;; (exec-results (cmd-run->list fullrunscript)) ;; (list ">" (conc test-name "-run.log"))))
;; (success exec-results)) ;; (eq? (cadr exec-results) 0)))
(debug:print 2 "Output from running " fullrunscript ", pid " (vector-ref exit-info 0) " in work area "
work-area ":\n====\n exit code " (vector-ref exit-info 2) "\n" "====\n")
;; (sqlite3:finalize! db)
|
| ︙ | | |
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
|
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
|
-
+
-
-
+
-
-
+
|
;;
;; <linkdir> - <target> - <testname> [ - <itempath> ]
;;
;; All log file links should be stored relative to the top of link path
;;
;; <target> - <testname> [ - <itempath> ]
;;
(define (create-work-area db run-id test-id test-src-path disk-path testname itemdat)
(define (create-work-area run-id run-info keyvals test-id test-src-path disk-path testname itemdat)
(let* ((run-info (cdb:remote-run db:get-run-info #f run-id))
(item-path (item-list->path itemdat))
(key-vals (cdb:remote-run db:get-key-vals #f run-id))
(target (string-intersperse key-vals "/"))
|
| ︙ | | |
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
|
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
|
-
+
-
+
+
-
+
-
-
-
-
-
-
-
+
+
+
+
+
+
+
+
|
;; (system (conc "ln -sf " test-path " " testlink)))
(if (directory? test-path)
(begin
(let* ((ovrcmd (let ((cmd (config-lookup *configdat* "setup" "testcopycmd")))
(if cmd
;; substitute the TEST_SRC_PATH and TEST_TARG_PATH
(string-substitute "TEST_TARG_PATH" test-path
(string-substitute "TEST_SRC_PATH" test-src-path cmd))
(string-substitute "TEST_SRC_PATH" test-src-path cmd #t) #t)
#f)))
(cmd (if ovrcmd
ovrcmd
(conc "rsync -av" (if (debug:debug-mode 1) "" "q") " " test-src-path "/ " test-path "/")))
(conc "rsync -av" (if (debug:debug-mode 1) "" "q") " " test-src-path "/ " test-path "/"
" >> " test-path "/mt_launch.log 2>> " test-path "/mt_launch.log")))
(status (system cmd)))
(if (not (eq? status 0))
(debug:print 2 "ERROR: problem with running \"" cmd "\"")))
(list lnkpathf lnkpath ))
(list #f #f))))
;; 1. look though disks list for disk with most space
;; 2. create run dir on disk, path name is meaningful
;; 3. create link from run dir to megatest runs area
;; 4. remotely run the test on allocated host
;; - could be ssh to host from hosts table (update regularly with load)
;; - could be netbatch
;; (launch-test db (cadr status) test-conf))
(define (launch-test db run-id runname test-conf keyvallst test-name test-path itemdat params)
(define (launch-test test-id run-id run-info keyvals runname test-conf test-name test-path itemdat params)
(change-directory *toppath*)
(alist->env-vars ;; consolidate this code with the code in megatest.scm for "-execute"
(list ;; (list "MT_TEST_RUN_DIR" work-area)
(list "MT_RUN_AREA_HOME" *toppath*)
(list "MT_TEST_NAME" test-name)
;; (list "MT_ITEM_INFO" (conc itemdat))
(list "MT_RUNNAME" runname)
;; (list "MT_TARGET" mt_target)
))
(let* ((useshell (config-lookup *configdat* "jobtools" "useshell"))
(launcher (config-lookup *configdat* "jobtools" "launcher"))
(runscript (config-lookup test-conf "setup" "runscript"))
(ezsteps (> (length (hash-table-ref/default test-conf "ezsteps" '())) 0)) ;; don't send all the steps, could be big
(diskspace (config-lookup test-conf "requirements" "diskspace"))
(memory (config-lookup test-conf "requirements" "memory"))
(hosts (config-lookup *configdat* "jobtools" "workhosts"))
(let* ((useshell (config-lookup *configdat* "jobtools" "useshell"))
(launcher (config-lookup *configdat* "jobtools" "launcher"))
(runscript (config-lookup test-conf "setup" "runscript"))
(ezsteps (> (length (hash-table-ref/default test-conf "ezsteps" '())) 0)) ;; don't send all the steps, could be big
(diskspace (config-lookup test-conf "requirements" "diskspace"))
(memory (config-lookup test-conf "requirements" "memory"))
(hosts (config-lookup *configdat* "jobtools" "workhosts"))
(remote-megatest (config-lookup *configdat* "setup" "executable"))
(run-time-limit (configf:lookup test-conf "requirements" "runtimelim"))
;; FIXME SOMEDAY: not good how this is so obtuse, this hack is to
;; allow running from dashboard. Extract the path
;; from the called megatest and convert dashboard
;; or dboard to megatest
(local-megatest (let* ((lm (car (argv)))
(dir (pathname-directory lm))
(exe (pathname-strip-directory lm)))
|
| ︙ | | |
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
|
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
|
-
+
-
+
+
-
+
|
(work-area #f)
(toptest-work-area #f) ;; for iterated tests the top test contains data relevant for all
(diskpath #f)
(cmdparms #f)
(fullcmd #f) ;; (define a (with-output-to-string (lambda ()(write x))))
(mt-bindir-path #f)
(item-path (item-list->path itemdat))
(test-id (cdb:remote-run db:get-test-id #f run-id test-name item-path))
;; (test-id (cdb:remote-run db:get-test-id #f run-id test-name item-path))
(testinfo (cdb:get-test-info-by-id *runremote* test-id))
(mt_target (string-intersperse (map cadr keyvallst) "/"))
(mt_target (string-intersperse (map cadr keyvals) "/"))
(debug-param (append (if (args:get-arg "-debug") (list "-debug" (args:get-arg "-debug")) '())
(if (args:get-arg "-logging")(list "-logging") '()))))
(setenv "MT_ITEMPATH" item-path)
(if hosts (set! hosts (string-split hosts)))
;; set the megatest to be called on the remote host
(if (not remote-megatest)(set! remote-megatest local-megatest)) ;; "megatest"))
(set! mt-bindir-path (pathname-directory remote-megatest))
(if launcher (set! launcher (string-split launcher)))
;; set up the run work area for this test
(set! diskpath (get-best-disk *configdat*))
(if diskpath
(let ((dat (open-run-close create-work-area db run-id test-id test-path diskpath test-name itemdat)))
(let ((dat (create-work-area run-id run-info keyvals test-id test-path diskpath test-name itemdat)))
(set! work-area (car dat))
(set! toptest-work-area (cadr dat))
(debug:print-info 2 "Using work area " work-area))
(begin
(set! work-area (conc test-path "/tmp_run"))
(create-directory work-area #t)
(debug:print 0 "WARNING: No disk work area specified - running in the test directory under tmp_run")))
|
| ︙ | | |
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
|
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
|
+
-
+
|
(list 'runscript runscript)
(list 'run-id run-id )
(list 'test-id test-id )
(list 'itemdat itemdat )
(list 'megatest remote-megatest)
(list 'ezsteps ezsteps)
(list 'target mt_target)
(list 'runtlim (if run-time-limit (common:hms-string->seconds run-time-limit) #f))
(list 'env-ovrd (hash-table-ref/default *configdat* "env-override" '()))
(list 'set-vars (if params (hash-table-ref/default params "-setvars" #f)))
(list 'runname runname)
(list 'mt-bindir-path mt-bindir-path))))))) ;; (string-intersperse keyvallst " "))))
(list 'mt-bindir-path mt-bindir-path)))))))
;; clean out step records from previous run if they exist
;; (debug:print-info 4 "FIXMEEEEE!!!! This can be removed some day, perhaps move all test records to the test db?")
;; (open-run-close db:delete-test-step-records db test-id)
(change-directory work-area) ;; so that log files from the launch process don't clutter the test dir
(tests:test-set-status! test-id "LAUNCHED" "n/a" #f #f) ;; (if launch-results launch-results "FAILED"))
(cond
((and launcher hosts) ;; must be using ssh hostname
|
| ︙ | | |
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
|
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
|
+
+
+
-
+
+
+
-
-
+
+
+
+
-
+
+
+
-
-
-
-
-
-
-
-
-
+
+
+
+
+
+
+
+
+
|
(hash-table-ref/default test-conf "pre-launch-env-overrides" '())))
(miscprevvals (alist->env-vars ;; consolidate this code with the code in megatest.scm for "-execute"
(append (list (list "MT_TEST_RUN_DIR" work-area)
(list "MT_TEST_NAME" test-name)
(list "MT_ITEM_INFO" (conc itemdat))
(list "MT_RUNNAME" runname)
(list "MT_TARGET" mt_target)
(list "MT_ITEMPATH" item-path)
)
itemdat)))
;; Launchwait defaults to true, must override it to turn off wait
(launchwait (if (equal? (configf:lookup *configdat* "setup" "launchwait") "no") #f #t))
(launch-results (apply cmd-run-with-stderr->list ;; cmd-run-proc-each-line
(launch-results (apply (if launchwait
cmd-run-with-stderr->list
process-run)
(if useshell
(string-intersperse fullcmd " ")
(car fullcmd))
;; conc
(cdr fullcmd))))) ;; launcher fullcmd)));; (apply cmd-run-proc-each-line launcher print fullcmd))) ;; (cmd-run->list fullcmd))
(cdr fullcmd)))))
(if (not launchwait) ;; give the OS a little time to allow the process to start
(thread-sleep! 0.01))
(with-output-to-file "mt_launch.log"
(lambda ()
(if (list? launch-results)
(apply print launch-results)))
(apply print launch-results)
(print "NOTE: launched \"" fullcmd "\"\n but did not wait for it to proceed. Add the following to megatest.config \n[setup]\nlaunchwait yes\n if you have problems with this"))
#:append))
(debug:print 2 "Launching completed, updating db")
(debug:print 2 "Launch results: " launch-results)
(if (not launch-results)
(begin
(print "ERROR: Failed to run " (string-intersperse fullcmd " ") ", exiting now")
;; (sqlite3:finalize! db)
;; good ole "exit" seems not to work
;; (_exit 9)
;; but this hack will work! Thanks go to Alan Post of the Chicken email list
;; NB// Is this still needed? Should be safe to go back to "exit" now?
(process-signal (current-process-id) signal/kill)
))
|