Megatest

Check-in [846b99e992]
Login
Overview
Comment:Trial work on exception handling for sqlite issues
Downloads: Tarball | ZIP archive | SQL archive
Timelines: family | ancestors | descendants | both | sqlite-trials
Files: files | file ages | folders
SHA1: 846b99e992aab1b8e971fa34eec6f58faba6a4e3
User & Date: mrwellan on 2014-03-26 22:17:20
Other Links: branch diff | manifest | tags
Context
2014-03-27
11:13
Tweaked timeouts and added more agressive exception handling to sqlite3 calls. Also cd to MT_RUN_AREA_HOME in dashboard if run in a test enviroment Closed-Leaf check-in: 158702d7aa user: mrwellan tags: sqlite-trials
2014-03-26
22:17
Trial work on exception handling for sqlite issues check-in: 846b99e992 user: mrwellan tags: sqlite-trials
2014-03-25
12:30
Bumped version. check-in: fd82e8252e user: icfadm tags: v1.55, v1.5516rc1
Changes

Modified db.scm from [820999b8e9] to [fc53c2f8f2].

71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
	    (exit))))
  (let* ((dbpath    (conc *toppath* "/megatest.db")) ;; fname)
	 (dbexists  (file-exists? dbpath))
	 (write-access (file-write-access? dbpath))
	 (db        (sqlite3:open-database dbpath)) ;; (never-give-up-open-db dbpath))
	 (handler   (make-busy-timeout (if (args:get-arg "-override-timeout")
					   (string->number (args:get-arg "-override-timeout"))
					   136000)))) ;; 136000))) ;; 136000 = 2.2 minutes
    (if (and dbexists
	     (not write-access))
	(set! *db-write-access* write-access)) ;; only unset so other db's also can use this control
    (debug:print-info 11 "open-db, dbpath=" dbpath " argv=" (argv))
    (if write-access (sqlite3:set-busy-handler! db handler))
    (if (not dbexists)
	(db:initialize db))







|







71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
	    (exit))))
  (let* ((dbpath    (conc *toppath* "/megatest.db")) ;; fname)
	 (dbexists  (file-exists? dbpath))
	 (write-access (file-write-access? dbpath))
	 (db        (sqlite3:open-database dbpath)) ;; (never-give-up-open-db dbpath))
	 (handler   (make-busy-timeout (if (args:get-arg "-override-timeout")
					   (string->number (args:get-arg "-override-timeout"))
					   30)))) ;; 136000))) ;; 136000 = 2.2 minutes
    (if (and dbexists
	     (not write-access))
	(set! *db-write-access* write-access)) ;; only unset so other db's also can use this control
    (debug:print-info 11 "open-db, dbpath=" dbpath " argv=" (argv))
    (if write-access (sqlite3:set-busy-handler! db handler))
    (if (not dbexists)
	(db:initialize db))
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
  (handle-exceptions
   exn
   (begin
     (debug:print 0 "EXCEPTION: database probably overloaded or unreadable.")
     (debug:print 0 "  " ((condition-property-accessor 'exn 'message) exn))
     (print-call-chain)
     (thread-sleep! (random 120))
     (debug:print-info 0 "trying db call one more time....")
     (apply open-run-close-no-exception-handling proc idb params))
   (apply open-run-close-no-exception-handling proc idb params)))

;; (define open-run-close open-run-close-exception-handling)
(define open-run-close open-run-close-no-exception-handling)

(define *global-delta* 0)
(define *last-global-delta-printed* 0)

(define (open-run-close-measure  proc idb . params)
  (debug:print-info 11 "open-run-close-measure START, idb=" idb ", params=" params)
  (let* ((start-ms (current-milliseconds))







|




|







108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
  (handle-exceptions
   exn
   (begin
     (debug:print 0 "EXCEPTION: database probably overloaded or unreadable.")
     (debug:print 0 "  " ((condition-property-accessor 'exn 'message) exn))
     (print-call-chain)
     (thread-sleep! (random 120))
     (debug:print-info 0 "trying db call one more time....this may never recover, if necessary kill process " (current-process-id) " on host " (current-host-name) " to clean up")
     (apply open-run-close-no-exception-handling proc idb params))
   (apply open-run-close-no-exception-handling proc idb params)))

;; (define open-run-close open-run-close-exception-handling)
(define open-run-close open-run-close-exception-handling)

(define *global-delta* 0)
(define *last-global-delta-printed* 0)

(define (open-run-close-measure  proc idb . params)
  (debug:print-info 11 "open-run-close-measure START, idb=" idb ", params=" params)
  (let* ((start-ms (current-milliseconds))
269
270
271
272
273
274
275

276
277
278
279
280
281
282
283
	   (debug:print 2 "ERROR: problem accessing test db " work-area ", you probably should clean and re-run this test"
			((condition-property-accessor 'exn 'message) exn))
	   (set! db (sqlite3:open-database ":memory:"))) ;; open an in-memory db to allow readonly access 
	 (set! db (sqlite3:open-database dbpath)))
	(if *db-write-access* (sqlite3:set-busy-handler! db handler))
	(if (not dbexists)
	    (begin

	      (sqlite3:execute db "PRAGMA synchronous = FULL;")
	      (debug:print-info 11 "Initialized test database " dbpath)
	      (db:testdb-initialize db)))
	;; (sqlite3:execute db "PRAGMA synchronous = 0;")
	(debug:print-info 11 "open-test-db END (sucessful)" work-area)
	;; now let's test that everything is correct
	(handle-exceptions
	 exn







>
|







269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
	   (debug:print 2 "ERROR: problem accessing test db " work-area ", you probably should clean and re-run this test"
			((condition-property-accessor 'exn 'message) exn))
	   (set! db (sqlite3:open-database ":memory:"))) ;; open an in-memory db to allow readonly access 
	 (set! db (sqlite3:open-database dbpath)))
	(if *db-write-access* (sqlite3:set-busy-handler! db handler))
	(if (not dbexists)
	    (begin
	      ;; Why use FULL here? This data is not that critical
	      ;; (sqlite3:execute db "PRAGMA synchronous = FULL;")
	      (debug:print-info 11 "Initialized test database " dbpath)
	      (db:testdb-initialize db)))
	;; (sqlite3:execute db "PRAGMA synchronous = 0;")
	(debug:print-info 11 "open-test-db END (sucessful)" work-area)
	;; now let's test that everything is correct
	(handle-exceptions
	 exn

Modified runs.scm from [16fd05bb4c] to [d043a7c00f].

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
             ;; (notinqueue (filter (lambda (x)
             ;;    		   (not (member x allinqueue)))
             ;;    		 prereqstrs))
	     (give-up    #f))

	;; We can get here when a prereq has not been run due to *it* having a prereq that failed.
	;; We need to use this to dequeue this item as CANNOTRUN


	(for-each (lambda (prereq)
		    (if (eq? (hash-table-ref/default test-registry prereq 'justfine) 'CANNOTRUN)
			(set! give-up #t)))
		  prereqstrs)

	(if (and give-up
		 (not (and (null? tal)(null? reg))))
	    (let ((trimmed-tal (mt:discard-blocked-tests run-id hed tal test-records))
		  (trimmed-reg (mt:discard-blocked-tests run-id hed reg test-records)))
	      (debug:print 1 "WARNING: test " hed " has discarded prerequisites, removing it from the queue")

	      (let ((test-id (cdb:remote-run db:get-test-id-cached #f run-id hed "")))
		(mt:test-set-state-status-by-id test-id "DEQUED" "PREQ_FAIL" "Failed to run due to failed prerequisites"))
	      
	      (if (and (null? trimmed-tal)
		       (null? trimmed-reg))
		  #f
		  (list (runs:queue-next-hed trimmed-tal trimmed-reg reglen regfull)
			(runs:queue-next-tal trimmed-tal trimmed-reg reglen regfull)
			(runs:queue-next-reg trimmed-tal trimmed-reg reglen regfull)







>
>
|
|
|
|
>







|







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
487
             ;; (notinqueue (filter (lambda (x)
             ;;    		   (not (member x allinqueue)))
             ;;    		 prereqstrs))
	     (give-up    #f))

	;; We can get here when a prereq has not been run due to *it* having a prereq that failed.
	;; We need to use this to dequeue this item as CANNOTRUN
	;; 
	(if (member testmode '(toplevel))
	    (for-each (lambda (prereq)
			(if (eq? (hash-table-ref/default test-registry prereq 'justfine) 'CANNOTRUN)
			    (set! give-up #t)))
		      prereqstrs))

	(if (and give-up
		 (not (and (null? tal)(null? reg))))
	    (let ((trimmed-tal (mt:discard-blocked-tests run-id hed tal test-records))
		  (trimmed-reg (mt:discard-blocked-tests run-id hed reg test-records)))
	      (debug:print 1 "WARNING: test " hed " has discarded prerequisites, removing it from the queue")

	      (let ((test-id (cdb:remote-run db:get-test-id-cached #f run-id hed "")))
		(mt:test-set-state-status-by-id test-id "DEQUEUED" "PREQ_FAIL" "Failed to run due to failed prerequisites"))
	      
	      (if (and (null? trimmed-tal)
		       (null? trimmed-reg))
		  #f
		  (list (runs:queue-next-hed trimmed-tal trimmed-reg reglen regfull)
			(runs:queue-next-tal trimmed-tal trimmed-reg reglen regfull)
			(runs:queue-next-reg trimmed-tal trimmed-reg reglen regfull)