Megatest

Check-in [f824fb4d79]
Login
Overview
Comment:added re-try exception handling to lock-queue
Downloads: Tarball | ZIP archive | SQL archive
Timelines: family | ancestors | descendants | both | v1.55
Files: files | file ages | folders
SHA1: f824fb4d79a979924a74c1f8a3666db039444487
User & Date: matt on 2014-07-22 00:10:00
Other Links: branch diff | manifest | tags
Context
2014-07-22
10:11
Changed KILLED status from FAIL to KILLED and treat INCOMPLETE as if it was COMPLETED check-in: a35313652a user: mrwellan tags: v1.55
00:10
added re-try exception handling to lock-queue check-in: f824fb4d79 user: matt tags: v1.55
2014-07-21
17:06
Added fix for corner case where no tests were runnable but the other conditions did not catch it. check-in: e5dba3e2c7 user: mrwellan tags: v1.55
Changes

Modified lock-queue.scm from [a9f4c5425b] to [0b444a0bb7].

19
20
21
22
23
24
25
26
27
28
29
30
31
32
33







34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
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
105
106
107
108
109







110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
(declare (uses common))

;;======================================================================
;; attempt to prevent overlapping updates of rollup files by queueing
;; update requests in an sqlite db
;;======================================================================

(define (lock-queue:open-db fname)
  (let* ((actualfname (conc fname ".lockdb"))
	 (dbexists (file-exists? actualfname))
	 (db       (sqlite3:open-database actualfname))
	 (handler  (make-busy-timeout 136000)))
    (if dbexists
	db
	(begin







	  (sqlite3:execute 
	   db
	   "CREATE TABLE IF NOT EXISTS queue (
  	      id         INTEGER PRIMARY KEY,
              test_id    INTEGER,
              start_time INTEGER,
              state      TEXT,
              CONSTRAINT queue_constraint UNIQUE (test_id));")
	  (sqlite3:execute
	   db
	   "CREATE TABLE IF NOT EXISTS runlocks (
              id         INTEGER PRIMARY KEY,
              test_id    INTEGER,
              run_lock   TEXT,
              CONSTRAINT runlock_constraint UNIQUE (run_lock));")))
    (sqlite3:set-busy-handler! db handler)
    db))

(define (lock-queue:set-state db test-id newstate)







  (sqlite3:execute db "UPDATE queue SET state=? WHERE test_id=?;"
		   newstate
		   test-id))

(define (lock-queue:any-younger? db mystart test-id)
  (let ((res #f))







    (sqlite3:for-each-row
     (lambda (tid)
       ;; Actually this should not be needed as mystart cannot be simultaneously less than and test-id same as 
       (if (not (equal? tid test-id)) 
	   (set! res tid)))
     db
     "SELECT test_id FROM queue WHERE start_time > ?;" mystart)
    res))

(define (lock-queue:get-lock db test-id)
  (let ((res       #f)
	(lckqry    (sqlite3:prepare db "SELECT test_id,run_lock FROM runlocks WHERE run_lock='locked';"))
	(mklckqry  (sqlite3:prepare db "INSERT INTO runlocks (test_id,run_lock) VALUES (?,'locked');")))
    (let ((result 
	   (handle-exceptions
	    exn




	    #f
	    (sqlite3:with-transaction
	     db
	     (lambda ()
	       (sqlite3:for-each-row (lambda (tid lockstate)
				       (set! res (list tid lockstate)))
				     lckqry)
	       (if res
		   (if (equal? (car res) test-id)
		       #t ;; already have the lock
		       #f)
		   (begin
		     (sqlite3:execute mklckqry test-id)
		     ;; if no error handled then return #t for got the lock
		     #t)))))))
      (sqlite3:finalize! lckqry)
      (sqlite3:finalize! mklckqry)
      result)))

(define (lock-queue:release-lock fname test-id)
  (let ((db (lock-queue:open-db fname)))







    (sqlite3:execute db "DELETE FROM runlocks WHERE test_id=?;" test-id)
    (sqlite3:finalize! db)))

(define (lock-queue:steal-lock db test-id)







  (sqlite3:execute db "DELETE FROM runlocks WHERE run_lock='locked';")
  (lock-queue:get-lock db test-it))

;; returns #f if ok to skip the task
;; returns #t if ok to proceed with task
;; otherwise waits
;;
(define (lock-queue:wait-turn fname test-id)
  (let ((db      (lock-queue:open-db fname))
	(mystart (current-seconds)))







    (sqlite3:execute
     db
     "INSERT OR REPLACE INTO queue (test_id,start_time,state) VALUES (?,?,'waiting');"
     test-id mystart)
    (thread-sleep! 1) ;; give other tests a chance to register
    (let ((result 
	   (let loop ((younger-waiting (lock-queue:any-younger? db mystart test-id)))
	     (if younger-waiting
		 (begin
		   ;; no need for us to wait. mark in the lock queue db as skipping
		   (lock-queue:set-state db test-id "skipping")
		   #f) ;; let the calling process know that nothing needs to be done
		 (if (lock-queue:get-lock db test-id)
		     #t
		     (if (> (- (current-seconds) mystart) 36000) ;; waited too long, steal the lock
			 (lock-queue:steal-lock db test-id)
			 (begin
			   (thread-sleep! 1)
			   (loop (lock-queue:any-younger? db mystart test-id)))))))))
      (sqlite3:finalize! db)
      result)))
	  
            
;; (use trace)
;; (trace lock-queue:get-lock lock-queue:release-lock lock-queue:wait-turn lock-queue:any-younger? lock-queue:set-state)







|







>
>
>
>
>
>
>
|
|
|





|
|
|



|



|
>
>
>
>
>
>
>
|
|
|

|

>
>
>
>
>
>
>
|
|
|
|
|
|
|
|

|






>
>
>
>
|


















|

>
>
>
>
>
>
>
|
|

|
>
>
>
>
>
>
>
|






|


>
>
>
>
>
>
>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|




19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
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
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
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
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
(declare (uses common))

;;======================================================================
;; attempt to prevent overlapping updates of rollup files by queueing
;; update requests in an sqlite db
;;======================================================================

(define (lock-queue:open-db fname #!key (count 10))
  (let* ((actualfname (conc fname ".lockdb"))
	 (dbexists (file-exists? actualfname))
	 (db       (sqlite3:open-database actualfname))
	 (handler  (make-busy-timeout 136000)))
    (if dbexists
	db
	(begin
	  (handle-exceptions
	   exn
	   (begin
	     (thread-sleep! 10)
	     (if (> count 0)
		 (lock-queue:open-db fname count: (- count 1))
		 db))
	   (sqlite3:execute 
	    db
	    "CREATE TABLE IF NOT EXISTS queue (
  	      id         INTEGER PRIMARY KEY,
              test_id    INTEGER,
              start_time INTEGER,
              state      TEXT,
              CONSTRAINT queue_constraint UNIQUE (test_id));")
	   (sqlite3:execute
	    db
	    "CREATE TABLE IF NOT EXISTS runlocks (
              id         INTEGER PRIMARY KEY,
              test_id    INTEGER,
              run_lock   TEXT,
              CONSTRAINT runlock_constraint UNIQUE (run_lock));"))))
    (sqlite3:set-busy-handler! db handler)
    db))

(define (lock-queue:set-state db test-id newstate #!key (count 10))
  (handle-exceptions
   exn
   (begin
     (thread-sleep! 10)
     (if (> count 0)
	 (lock-queue:set-state db test-id newstate (- count 1))
	 #f))
   (sqlite3:execute db "UPDATE queue SET state=? WHERE test_id=?;"
		    newstate
		    test-id)))

(define (lock-queue:any-younger? db mystart test-id #!key (count 10))
  (let ((res #f))
    (handle-exceptions
     exn
     (begin
       (thread-sleep! 10)
       (if (> count 0)
	   (lock-queue:any-younger? db mystart test-id count: (- count 1))
	   #f))
     (sqlite3:for-each-row
      (lambda (tid)
	;; Actually this should not be needed as mystart cannot be simultaneously less than and test-id same as 
	(if (not (equal? tid test-id)) 
	    (set! res tid)))
      db
      "SELECT test_id FROM queue WHERE start_time > ?;" mystart)
     res)))

(define (lock-queue:get-lock db test-id #!key (count 10))
  (let ((res       #f)
	(lckqry    (sqlite3:prepare db "SELECT test_id,run_lock FROM runlocks WHERE run_lock='locked';"))
	(mklckqry  (sqlite3:prepare db "INSERT INTO runlocks (test_id,run_lock) VALUES (?,'locked');")))
    (let ((result 
	   (handle-exceptions
	    exn
	    (begin
	      (thread-sleep! 10)
	      (if (> count 0)
		  (lock-queue:get-lock db test-id count: (- count 1)))
	      #f)
	    (sqlite3:with-transaction
	     db
	     (lambda ()
	       (sqlite3:for-each-row (lambda (tid lockstate)
				       (set! res (list tid lockstate)))
				     lckqry)
	       (if res
		   (if (equal? (car res) test-id)
		       #t ;; already have the lock
		       #f)
		   (begin
		     (sqlite3:execute mklckqry test-id)
		     ;; if no error handled then return #t for got the lock
		     #t)))))))
      (sqlite3:finalize! lckqry)
      (sqlite3:finalize! mklckqry)
      result)))

(define (lock-queue:release-lock fname test-id #!key (count 10))
  (let ((db (lock-queue:open-db fname)))
    (handle-exceptions
     exn
     (begin
       (thread-sleep! 10)
       (if (> count 0)
	   (lock-queue:release-lock fname test-id count: (- count 1))
	   #f))
     (sqlite3:execute db "DELETE FROM runlocks WHERE test_id=?;" test-id)
     (sqlite3:finalize! db))))

(define (lock-queue:steal-lock db test-id #!key (count 10))
  (handle-exceptions
   exn
   (begin
     (thread-sleep! 10)
     (if (> count 0)
	 (lock-queue:steal-lock db test-id count: (- count 1))
	 #f))
   (sqlite3:execute db "DELETE FROM runlocks WHERE run_lock='locked';"))
  (lock-queue:get-lock db test-it))

;; returns #f if ok to skip the task
;; returns #t if ok to proceed with task
;; otherwise waits
;;
(define (lock-queue:wait-turn fname test-id #!key (count 10))
  (let ((db      (lock-queue:open-db fname))
	(mystart (current-seconds)))
    (handle-exceptions
     exn
     (begin
       (thread-sleep! 10)
       (if (> count 0)
	   (lock-queue:wait-turn fname test-id count: (- count 1))
	   #f))
     (sqlite3:execute
      db
      "INSERT OR REPLACE INTO queue (test_id,start_time,state) VALUES (?,?,'waiting');"
      test-id mystart)
     (thread-sleep! 1) ;; give other tests a chance to register
     (let ((result 
	    (let loop ((younger-waiting (lock-queue:any-younger? db mystart test-id)))
	      (if younger-waiting
		  (begin
		    ;; no need for us to wait. mark in the lock queue db as skipping
		    (lock-queue:set-state db test-id "skipping")
		    #f) ;; let the calling process know that nothing needs to be done
		  (if (lock-queue:get-lock db test-id)
		      #t
		      (if (> (- (current-seconds) mystart) 36000) ;; waited too long, steal the lock
			  (lock-queue:steal-lock db test-id)
			  (begin
			    (thread-sleep! 1)
			    (loop (lock-queue:any-younger? db mystart test-id)))))))))
       (sqlite3:finalize! db)
       result))))
	  
            
;; (use trace)
;; (trace lock-queue:get-lock lock-queue:release-lock lock-queue:wait-turn lock-queue:any-younger? lock-queue:set-state)

tests/installall/config/megatest.config.dat became a regular file with contents [736a5da885].

tests/installall/config/runconfigs.config.dat became a regular file with contents [3b8f260acb].