Megatest

Check-in [31896202a9]
Login
Overview
Comment:Break up find and mark incompletes into read portion and write portion - prevents unnecessarily starting servers
Downloads: Tarball | ZIP archive | SQL archive
Timelines: family | ancestors | descendants | both | archiving | acatqa-bug-introduced
Files: files | file ages | folders
SHA1: 31896202a9e53d229dffd6c61c20c8acfe897df5
User & Date: matt on 2014-12-02 22:48:53
Other Links: branch diff | manifest | tags
Context
2014-12-02
23:19
Added small random delay in server lock. Fix bad params in rmt call in runs.scm check-in: bcc2bef4ad user: matt tags: archiving
22:48
Break up find and mark incompletes into read portion and write portion - prevents unnecessarily starting servers check-in: 31896202a9 user: matt tags: archiving, acatqa-bug-introduced
22:16
Pass through read-only queries but force server on write check-in: c9ccc57c3f user: matt tags: archiving
Changes

Modified api.scm from [a72013393c] to [fafaf8a322].

43
44
45
46
47
48
49
50



51
52
53
54
55
56
57
43
44
45
46
47
48
49

50
51
52
53
54
55
56
57
58
59







-
+
+
+







    get-runs
    get-all-run-ids
    get-prev-run-ids
    get-run-ids-matching-target
    get-runs-by-patt
    get-steps-data
    login
    testmeta-get-record))
    testmeta-get-record
    have-incompletes?
    ))

(define api:write-queries
  '(
    ;; SERVERS
    start-server
    kill-server

66
67
68
69
70
71
72
73

74
75
76
77
78
79
80
68
69
70
71
72
73
74

75
76
77
78
79
80
81
82







-
+








    ;; RUNS
    register-run
    set-tests-state-status
    delete-run
    lock/unlock-run
    update-run-event_time
    find-and-mark-incomplete
    mark-incomplete

    ;; STEPS
    teststep-set-status!

    ;; TEST DATA
    test-data-rollup
    csv->test-data
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
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







-
-











+








	    ;; RUNS
	    ((register-run)                 (apply db:register-run dbstruct params))
	    ((set-tests-state-status)       (apply db:set-tests-state-status dbstruct params))
	    ((delete-run)                   (apply db:delete-run dbstruct params))
	    ((lock/unlock-run)              (apply db:lock/unlock-run dbstruct params))
	    ((update-run-event_time)        (apply db:update-run-event_time dbstruct params))
	    ((find-and-mark-incomplete)     (apply db:find-and-mark-incomplete dbstruct params))


	    ;; STEPS
	    ((teststep-set-status!)         (apply db:teststep-set-status! dbstruct params))

	    ;; TEST DATA
	    ((test-data-rollup)             (apply db:test-data-rollup dbstruct params))
	    ((csv->test-data)               (apply db:csv->test-data dbstruct params))

	    ;; MISC
	    ((sync-inmem->db)               (let ((run-id (car params)))
					      (db:sync-touched dbstruct run-id force-sync: #t)))
	    ((mark-incomplete)              (apply db:find-and-mark-incomplete dbstruct params))

	    ;; TESTMETA
	    ((testmeta-add-record)       (apply db:testmeta-add-record dbstruct params))
	    ((testmeta-update-field)     (apply db:testmeta-update-field dbstruct params))

	    ;; TASKS
	    ((tasks-add)                 (apply tasks:add dbstruct params))   
197
198
199
200
201
202
203

204
205
206
207
208
209
210
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212







+







	    ((get-runs-by-patt)             (apply db:get-runs-by-patt dbstruct params))
	    ((get-run-name-from-id)         (apply db:get-run-name-from-id dbstruct params))

	    ;; STEPS
	    ((get-steps-data)               (apply db:get-steps-data dbstruct params))

	    ;; MISC
	    ((have-incompletes?)            (apply db:have-incompletes? dbstruct params))
	    ((login)                        (apply db:login dbstruct params))
	    ((general-call)                 (let ((stmtname   (car params))
						  (run-id     (cadr params))
						  (realparams (cddr params)))
					      (db:with-db dbstruct run-id #t ;; these are all for modifying the db
							  (lambda (db)
							    (db:general-call db stmtname realparams)))))

Modified db.scm from [845689c9c4] to [d9d867f59c].

933
934
935
936
937
938
939






















































940
941
942
943
944
945
946
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000







+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+







;;======================================================================
;; D B   U T I L S
;;======================================================================

;;======================================================================
;; M A I N T E N A N C E
;;======================================================================

(define (db:have-incompletes? dbstruct run-id ovr-deadtime)
  (let* ((dbdat        (db:get-db dbstruct run-id))
	 (db           (db:dbdat-get-db dbdat))
	 (incompleted '())
	 (oldlaunched '())
	 (toplevels   '())
	 (deadtime-str (configf:lookup *configdat* "setup" "deadtime"))
	 (deadtime     (if (and deadtime-str
				(string->number deadtime-str))
			   (string->number deadtime-str)
			   7200))) ;; two hours
    (if (number? ovr-deadtime)(set! deadtime ovr-deadtime))
    
    ;; in RUNNING or REMOTEHOSTSTART for more than 10 minutes
    ;;
    ;; HOWEVER: this code in run:test seems to work fine
    ;;              (> (- (current-seconds)(+ (db:test-get-event_time testdat)
    ;;                     (db:test-get-run_duration testdat)))
    ;;                    600) 
    (db:delay-if-busy dbdat)
    (sqlite3:for-each-row 
     (lambda (test-id run-dir uname testname item-path)
       (if (and (equal? uname "n/a")
		(equal? item-path "")) ;; this is a toplevel test
	   ;; what to do with toplevel? call rollup?
	   (begin
	     (set! toplevels   (cons (list test-id run-dir uname testname item-path run-id) toplevels))
	     (debug:print-info 0 "Found old toplevel test in RUNNING state, test-id=" test-id))
	   (set! incompleted (cons (list test-id run-dir uname testname item-path run-id) incompleted))))
     db
     "SELECT id,rundir,uname,testname,item_path FROM tests WHERE run_id=? AND (strftime('%s','now') - event_time) > (run_duration + ?) AND state IN ('RUNNING','REMOTEHOSTSTART');"
     run-id deadtime)

    ;; in LAUNCHED for more than one day. Could be long due to job queues TODO/BUG: Need override for this in config
    ;;
    (db:delay-if-busy dbdat)
    (sqlite3:for-each-row
     (lambda (test-id run-dir uname testname item-path)
       (if (and (equal? uname "n/a")
		(equal? item-path "")) ;; this is a toplevel test
	   ;; what to do with toplevel? call rollup?
	   (set! toplevels   (cons (list test-id run-dir uname testname item-path run-id) toplevels))
	   (set! oldlaunched (cons (list test-id run-dir uname testname item-path run-id) oldlaunched))))
     db
     "SELECT id,rundir,uname,testname,item_path FROM tests WHERE run_id=? AND (strftime('%s','now') - event_time) > 86400 AND state IN ('LAUNCHED');"
     run-id)
    
    (debug:print-info 18 "Found " (length oldlaunched) " old LAUNCHED items, " (length toplevels) " old LAUNCHED toplevel tests and " (length incompleted) " tests marked RUNNING but apparently dead.")
    (if (and (null? incompleted)
	     (null? oldlaunched)
	     (null? toplevels))
	#f
	#t)))

;;  select end_time-now from
;;      (select testname,item_path,event_time+run_duration as
;;                          end_time,strftime('%s','now') as now from tests where state in
;;      ('RUNNING','REMOTEHOSTSTART','LAUNCED'));

(define (db:find-and-mark-incomplete dbstruct run-id ovr-deadtime)

Modified rmt.scm from [8176d4a91d] to [43798d972b].

515
516
517
518
519
520
521

522

523
524
525
526
527
528
529
515
516
517
518
519
520
521
522

523
524
525
526
527
528
529
530







+
-
+







(define (rmt:update-run-event_time run-id)
  (rmt:send-receive 'update-run-event_time #f (list run-id)))

(define (rmt:get-runs-by-patt  keys runnamepatt targpatt offset limit)
  (rmt:send-receive 'get-runs-by-patt #f (list keys runnamepatt targpatt offset limit)))

(define (rmt:find-and-mark-incomplete run-id ovr-deadtime)
  (if (rmt:send-receive 'have-incompletes? run-id (list run-id ovr-deadtime))
  (rmt:send-receive 'find-and-mark-incomplete run-id (list run-id ovr-deadtime)))
      (rmt:send-receive 'mark-incomplete run-id (list run-id ovr-deadtime))))

;;======================================================================
;; M U L T I R U N   Q U E R I E S
;;======================================================================

;; Need to move this to multi-run section and make associated changes
(define (rmt:find-and-mark-incomplete-all-runs #!key (ovr-deadtime #f))