Megatest

Check-in [41eeb0e45f]
Login
☰Home Timeline Files Branches Tags Tickets Wiki Login
Overview
Comment:Added speculative dynamically calculated delay to reduce db hammering
Downloads: Tarball | ZIP archive | SQL archive
Timelines: family | ancestors | descendants | both | v1.65
Files: files | file ages | folders
SHA1: 41eeb0e45fe329b8380a77cd2c27abb7c5bee731
User & Date: matt on 2020-10-17 00:46:02
Other Links: branch diff | manifest | tags
Context
2020-10-17
23:06
Merged in side2 check-in: a33d7eb0be user: matt tags: v1.65
14:35
Try query rest Leaf check-in: dea5af2fba user: matt tags: v1.65-side3
00:46
Added speculative dynamically calculated delay to reduce db hammering check-in: 41eeb0e45f user: matt tags: v1.65
2020-10-16
18:09
variable propogation fix check-in: b156cc02a8 user: pjhatwal tags: v1.65
Changes
Hide Diffs Side-by-Side Diffs Ignore Whitespace Patch

Modified db.scm from [fb3a18f52f] to [055cfbeff9].

︙
3213
3214
3215
3216
3217
3218
3219
3220

3221
3222
3223
3224
3225
3226
3227
3228
3229
3230








3231
3232
3233
3234
3235
3236
3237
3238
3239
3240
3241
3242
3243
3244
3245
3246
3247
3248
3249
3250

3251
3252
3253

3254
3255
3256
3257
3258
3259
3260

3213
3214
3215
3216
3217
3218
3219

3220










3221
3222
3223
3224
3225
3226
3227
3228
3229
3230
3231
3232
3233
3234
3235
3236
3237
3238
3239
3240
3241
3242
3243
3244
3245
3246
3247

3248



3249
3250
3251
3252
3253
3254
3255
3256








-
+
-
-
-
-
-
-
-
-
-
-
+
+
+
+
+
+
+
+



















-
+
-
-
-
+








       (if newstatus  (sqlite3:execute db "UPDATE tests SET status=?  WHERE id=?;" newstatus  test-id))
       (if newcomment (sqlite3:execute db "UPDATE tests SET comment=? WHERE id=?;" newcomment ;; (sdb:qry 'getid newcomment)
				       test-id))))))
  (mt:process-triggers dbstruct run-id test-id newstate newstatus))

;; NEW BEHAVIOR: Count tests running in all runs!
;;
(define (db:get-count-tests-running dbstruct run-id fastmode)
(define (db:get-count-tests-running dbstruct run-id)
  (let* ((qry (if fastmode
		  "SELECT count(id) FROM tests WHERE state in ('RUNNING','LAUNCHED','REMOTEHOSTSTART') AND NOT (uname = 'n/a' AND item_path = '') LIMIT 1;"
		  "SELECT count(id) FROM tests WHERE state in ('RUNNING','LAUNCHED','REMOTEHOSTSTART') AND NOT (uname = 'n/a' AND item_path = '');")))
  (db:with-db
   dbstruct
   run-id
   #f
   (lambda (db)
     (let* ((stmth (db:get-cache-stmth dbstruct  db qry)))
       (sqlite3:first-result stmth))))))
  (let* ((qry  "SELECT count(id) FROM tests WHERE state in ('RUNNING','LAUNCHED','REMOTEHOSTSTART') AND NOT (uname = 'n/a' AND item_path = '');"))
    (db:with-db
     dbstruct
     run-id
     #f
     (lambda (db)
       (let* ((stmth (db:get-cache-stmth dbstruct db qry)))
	 (sqlite3:first-result stmth))))))

;; NEW BEHAVIOR: Count tests running in only one run!
;;
(define (db:get-count-tests-actually-running dbstruct run-id)
  (db:with-db
   dbstruct
   run-id
   #f
   (lambda (db)
     (sqlite3:first-result
      db
      ;; WARNING BUG EDIT ME - merged from v1.55 - not sure what is right here ...
      ;; "SELECT count(id) FROM tests WHERE state in ('RUNNING','LAUNCHED','REMOTEHOSTSTART') AND run_id NOT IN (SELECT id FROM runs WHERE state='deleted') AND NOT (uname = 'n/a' AND item_path = '');")
      "SELECT count(id) FROM tests WHERE state in ('RUNNING','REMOTEHOSTSTART','LAUNCHED') AND run_id=?;" 
      run-id)))) ;; NOT IN (SELECT id FROM runs WHERE state='deleted');")

;; NEW BEHAVIOR: Look only at single run with run-id
;; 
;; (define (db:get-running-stats dbstruct run-id)
(define (db:get-count-tests-running-for-run-id dbstruct run-id fastmode)
(define (db:get-count-tests-running-for-run-id dbstruct run-id)
  (let* ((qry (if fastmode
		  "SELECT count(id) FROM tests WHERE state in ('RUNNING','LAUNCHED','REMOTEHOSTSTART') AND run_id=? LIMIT 1;"
		  "SELECT count(id) FROM tests WHERE state in ('RUNNING','LAUNCHED','REMOTEHOSTSTART') AND run_id=?;")))
  (let* ((qry "SELECT count(id) FROM tests WHERE state in ('RUNNING','LAUNCHED','REMOTEHOSTSTART') AND run_id=?;"))
    (db:with-db
     dbstruct
     run-id
     #f
     (lambda (db)
       (let* ((stmth (db:get-cache-stmth dbstruct db qry)))
	 (sqlite3:first-result stmth run-id))))))
︙

Modified launch.scm from [d0067277fa] to [e8093b3e63].

︙
729
730
731
732
733
734
735
736

737
738
739
740
741
742
743

729
730
731
732
733
734
735

736
737
738
739
740
741
742
743








-
+








;; 100% COMPLETED/ (PASS,FAIL,ABORT etc.) ==> COMPLETED / X where X is same as itemized rollup
;; > 3 RUNNING with not test_dead do nothing (run should already be RUNNING/ na
;; > 0 RUNNING and test_dead then send KILLREQ ==> COMPLETED
;; 0 RUNNING ==> this is actually the first condition, should not get here

(define (launch:end-of-run-check run-id )
    (let* ((not-completed-cnt (rmt:get-not-completed-cnt run-id))  
           (running-cnt (rmt:get-count-tests-running-for-run-id run-id #f)) ;; fastmode=no
           (running-cnt (rmt:get-count-tests-running-for-run-id run-id))
           (all-test-launched (rmt:get-var (conc "lunch-complete-" run-id)))
           (current-state (rmt:get-run-state run-id))
           (current-status (rmt:get-run-status run-id)))
     ;;get-vars run-id to query metadata table to check if all completed. if all-test-launched = yes then only not-completed-cnt = 0 means everyting is completed if no entry found in the table do nothing 
     (debug:print 0 *default-log-port* "Running test cnt :" running-cnt)                      
     (rmt:set-state-status-and-roll-up-run  run-id current-state current-status)
     (runs:update-junit-test-reporter-xml run-id)
︙

Modified rmt.scm from [39d97c528a] to [342d9c7e08].

︙
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

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








-
-
+
+
+
+







-
-
+
+








	   (map (lambda (run-id)
		  (rmt:send-receive 'test-get-paths-matching-keynames-target-new run-id (list run-id keynames target res testpatt statepatt statuspatt runname)))
	   run-ids))))

(define (rmt:get-prereqs-not-met run-id waitons ref-test-name ref-item-path #!key (mode '(normal))(itemmaps #f))
  (rmt:send-receive 'get-prereqs-not-met run-id (list run-id waitons ref-test-name ref-item-path mode itemmaps)))

(define (rmt:get-count-tests-running-for-run-id run-id fastmode)
  (rmt:send-receive 'get-count-tests-running-for-run-id run-id (list run-id fastmode)))
(define (rmt:get-count-tests-running-for-run-id run-id)
  (if (number? run-id)
      (rmt:send-receive 'get-count-tests-running-for-run-id run-id (list run-id))
      0))

(define (rmt:get-not-completed-cnt run-id)
  (rmt:send-receive 'get-not-completed-cnt run-id (list run-id)))


;; Statistical queries

(define (rmt:get-count-tests-running run-id fastmode)
  (rmt:send-receive 'get-count-tests-running run-id (list run-id fastmode)))
(define (rmt:get-count-tests-running run-id)
  (rmt:send-receive 'get-count-tests-running run-id (list run-id)))

(define (rmt:get-count-tests-running-for-testname run-id testname)
  (rmt:send-receive 'get-count-tests-running-for-testname run-id (list run-id testname)))

(define (rmt:get-count-tests-running-in-jobgroup run-id jobgroup)
  (rmt:send-receive 'get-count-tests-running-in-jobgroup run-id (list run-id jobgroup)))
︙

Modified runs.scm from [030b929939] to [24707790d1].

︙
59
60
61
62
63
64
65
66




























67
68
69
70
71
72
73

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








-
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+








  (last-jobs-check-time    0)
  )

(defstruct runs:testdat
  hed tal reg reruns  test-record
  test-name item-path jobgroup
  waitons testmode  newtal itemmaps prereqs-not-met)
  

(module runsmod
    (
     runs:wait-if-seen-recently
     )
  
(import scheme chicken data-structures extras files)
(import posix typed-records srfi-18 srfi-69
	  md5 message-digest
	  regex srfi-1)

(define *last-seen-ht* (make-hash-table))

(define (runs:wait-if-seen-recently wait-until . keys)
  (let* ((full-key   (string-intersperse keys "-"))
	 (last-seen  (hash-table-ref/default *last-seen-ht* full-key 0))
	 (now        (current-seconds))
	 (delta      (- now last-seen))
	 (needed     (if (< delta wait-until)
			 0
			 (- wait-until delta))))
    (if (> needed 0)(thread-sleep! needed))
    (hash-table-set! *last-seen-ht* full-key (current-seconds))
    needed))
)

(import runsmod)
    
;; look in the $MT_RUN_AREA_HOME/.softlocks directory for key-host-pid.softlock files
;;  - remove any that are over 3600 seconds old
;;  - if there are any that are younger than 10 seconds
;;      * sleep 10 seconds
;;      * touch my key-host-pid.softlock file
;;      * return
;;  - if there are no files younger than 10 seconds
︙
319
320
321
322
323
324
325
326

327
328
329
330
331
332
333

346
347
348
349
350
351
352

353
354
355
356
357
358
359
360








-
+








  ;; Take advantage of a good place to exit if running the one-pass methodology
  (if (and (> (runs:dat-can-run-more-tests-count runsdat) 20)
	   (args:get-arg "-one-pass"))
      (exit 0))

  (if (runs:dat-load-mgmt-function runsdat)((runs:dat-load-mgmt-function runsdat)))

  (let* ((num-running             (rmt:get-count-tests-running run-id #f)) ;; fastmode=no
  (let* ((num-running             (rmt:get-count-tests-running run-id)) 
	 (num-running-in-jobgroup (rmt:get-count-tests-running-in-jobgroup run-id jobgroup))
	 (job-group-limit         (let ((jobg-count (configf:lookup *configdat* "jobgroups" jobgroup)))
				    (if (string? jobg-count)
					(string->number jobg-count)
					jobg-count))))
    (if (> (+ num-running num-running-in-jobgroup) 0)
	(runs:inc-can-run-more-tests-count runsdat)) ;; (set! *runs:can-run-more-tests-count* (+ *runs:can-run-more-tests-count* 1)))
︙
1563
1564
1565
1566
1567
1568
1569
1570

1571
1572
1573
1574
1575
1576
1577

1590
1591
1592
1593
1594
1595
1596

1597
1598
1599
1600
1601
1602
1603
1604








-
+








				       ))
				   extras)
				  extras)
				'())))
	     (waitons     (delete-duplicates (append (tests:testqueue-get-waitons test-record) extra-waits) equal?))
	     (newtal      (append tal (list hed)))
	     (regfull     (>= (length reg) reglen))
	     (num-running (rmt:get-count-tests-running-for-run-id run-id #t)) ;; fastmode=yes
	     (num-running (rmt:get-count-tests-running-for-run-id run-id))
	     (testdat     (make-runs:testdat
			   hed: hed
			   tal: tal
			   reg: reg
			   reruns: reruns
			   test-record: test-record
			   test-name:   test-name
︙
1713
1714
1715
1716
1717
1718
1719



1720
1721
1722
1723
1724
1725
1726

1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756








+
+
+








				  ;; wait for load here
				  (if (runs:dat-load-mgmt-function runsdat)((runs:dat-load-mgmt-function runsdat)))
				  (loop-can-run-more (runs:can-run-more-tests runsdat run-id jobgroup max-concurrent-jobs)
						     (- remtries 1)))))))
		       )))))

	  ;; I'm not clear on why prereqs are gathered here TODO: verfiy this is needed
	  (let ((waited (runs:wait-if-seen-recently 5 "prereqs-not-met" hed item-path))) ;; if we've been down this path in the past 5 seconds - wait out the difference
	    (if (> waited 0)(debug:print 0 *default-log-port* "Waited for prereqs-not-met-"hed"-"item-path" for " waited "seconds.")))
	  
	  (runs:testdat-prereqs-not-met-set! testdat (rmt:get-prereqs-not-met run-id waitons hed item-path mode: testmode itemmaps: itemmaps))

	  ;; I'm not clear on why we'd capture running job counts here TODO: verify this is needed
	  (runs:dat-can-run-more-tests-set! runsdat (runs:can-run-more-tests runsdat run-id jobgroup max-concurrent-jobs))

	  (let ((loop-list (runs:process-expanded-tests runsdat testdat))) ;; in process-expanded-tests ultimately run:test -> launch-test -> test actually running
            (if loop-list (apply loop loop-list))))
︙
1829
1830
1831
1832
1833
1834
1835
1836

1837
1838
1839
1840
1841
1842
1843
1844
1845
1846

1847
1848
1849
1850
1851
1852
1853
1854
1855
1856

1857
1858
1859
1860
1861
1862
1863

1859
1860
1861
1862
1863
1864
1865

1866
1867
1868
1869
1870
1871
1872
1873
1874
1875

1876
1877
1878
1879
1880
1881
1882
1883
1884
1885

1886
1887
1888
1889
1890
1891
1892
1893








-
+









-
+









-
+








    (rmt:set-var (conc "lunch-complete-" run-id) "yes")  
        
    ;; now *if* -run-wait we wait for all tests to be done
    ;; Now wait for any RUNNING tests to complete (if in run-wait mode)
    ;; (if (runs:dat-load-mgmt-function runsdat)((runs:dat-load-mgmt-function runsdat)))
    (thread-sleep! 10) ;; I think there is a race condition here. Let states/statuses settle
    
    (let wait-loop ((num-running      (rmt:get-count-tests-running-for-run-id run-id #t)) ;; fastmode=yes
    (let wait-loop ((num-running      (rmt:get-count-tests-running-for-run-id run-id))
		    (prev-num-running 0))
      ;; (debug:print-info 13 *default-log-port* "num-running=" num-running ", prev-num-running=" prev-num-running)
      (if (and (or (args:get-arg "-run-wait")
		   (equal? (configf:lookup *configdat* "setup" "run-wait") "yes"))
	       (> num-running 0))
	  (begin
	    ;; Here we mark any old defunct tests as incomplete. Do this every fifteen minutes
	    ;; (debug:print 0 *default-log-port* "Got here eh! num-running=" num-running " (> num-running 0) " (> num-running 0))
	    (if (> (current-seconds)(+ last-time-incomplete 900))
		(let ((actual-num-running (rmt:get-count-tests-running-for-run-id run-id #f))) ;; fastmode=no
		(let ((actual-num-running (rmt:get-count-tests-running-for-run-id run-id)))
		  (debug:print-info 0 *default-log-port* "Marking stuck tests as INCOMPLETE while waiting for run " run-id
				    ". Running as pid " (current-process-id) " on " (get-host-name))
		  (set! last-time-incomplete (current-seconds)) ;; FIXME, this might be causing slow down - use of set!
		  (rmt:find-and-mark-incomplete run-id #f)
		  (debug:print-info 0 *default-log-port* "run-wait specified, waiting on " actual-num-running
				    " tests in RUNNING, REMOTEHOSTSTART or LAUNCHED state at "
				    (time->string (seconds->local-time (current-seconds))))))
	    ;; (if (runs:dat-load-mgmt-function runsdat)((runs:dat-load-mgmt-function runsdat)))
	    (thread-sleep! 5) ;; (if (>= num-running max-concurrent-jobs) 5 1))
	    (wait-loop (rmt:get-count-tests-running-for-run-id run-id #t) ;; fastmode=yes
	    (wait-loop (rmt:get-count-tests-running-for-run-id run-id)
		       num-running))))
    ;; LET* ((test-record
    ;; we get here on "drop through". All done!
    ;; this is moved to runs:run-testes since this function is getting called twice to ensure everthing is completed. 
    ;; (debug:print-info 0 *default-log-port* "Calling Post Hook")    
    ;; (runs:run-post-hook run-id)
    (debug:print-info 1 *default-log-port* "All tests launched")))
︙

This page was generated in about 0.141s by Fossil 2.24 [452f8e1ffc] 2024-03-26 15:36:00