Megatest

Check-in [a953799c68]
Login
Overview
Comment:Added special handling of errors when sync fails in a server. Catch errors if creating the link tree directory fails.
Downloads: Tarball | ZIP archive | SQL archive
Timelines: family | ancestors | descendants | both | v1.60
Files: files | file ages | folders
SHA1: a953799c682fda015f2cea4e9f928d7c72321db7
User & Date: matt on 2014-12-07 15:50:05
Other Links: branch diff | manifest | tags
Context
2014-12-07
18:47
Added caching of megatest.config, no locking yet... check-in: 800fea92da user: matt tags: v1.60
15:50
Added special handling of errors when sync fails in a server. Catch errors if creating the link tree directory fails. check-in: a953799c68 user: matt tags: v1.60
2014-12-04
22:23
Merged fixes back to v1.60 from archiving check-in: 33d8665511 user: matt tags: v1.60
Changes

Modified db.scm from [d9d867f59c] to [b058afcfaa].

491
492
493
494
495
496
497
498



499


500
501
502
503
504
505
506
507
     (debug:print 0 " message: " ((condition-property-accessor 'exn 'message) exn))
     (print "exn=" (condition->list exn))
     (debug:print 0 " status:  " ((condition-property-accessor 'sqlite3 'status) exn))
     (debug:print 0 " src db:  " (db:dbdat-get-path fromdb))
     (for-each (lambda (dbdat)
		 (debug:print 0 " dbpath:  " (db:dbdat-get-path dbdat)))
	       (cons todb slave-dbs))
     (if *server-run* ;; we are inside a server



	 (set! *time-to-exit* #t) ;; let watch dog know that it is time to die.


	 (exit 1)))
   (cond
    ((not fromdb) (debug:print 3 "WARNING: db:sync-tables called with fromdb missing") -1)
    ((not todb)   (debug:print 3 "WARNING: db:sync-tables called with todb missing") -2)
    ((not (sqlite3:database? (db:dbdat-get-db fromdb)))
     (debug:print 0 "ERROR: db:sync-tables called with fromdb not a database " fromdb) -3)
    ((not (sqlite3:database? (db:dbdat-get-db todb)))
     (debug:print 0 "ERROR: db:sync-tables called with todb not a database " todb) -4)







|
>
>
>
|
>
>
|







491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
     (debug:print 0 " message: " ((condition-property-accessor 'exn 'message) exn))
     (print "exn=" (condition->list exn))
     (debug:print 0 " status:  " ((condition-property-accessor 'sqlite3 'status) exn))
     (debug:print 0 " src db:  " (db:dbdat-get-path fromdb))
     (for-each (lambda (dbdat)
		 (debug:print 0 " dbpath:  " (db:dbdat-get-path dbdat)))
	       (cons todb slave-dbs))
     (if *server-run* ;; we are inside a server, throw a sync-failed error
	 (signal (make-composite-condition
		 (make-property-condition 'sync-failed 'message "db:sync-tables failed in a server context.")))))

	 ;; (set! *time-to-exit* #t) ;; let watch dog know that it is time to die.
	 ;; (tasks:server-set-state! (db:delay-if-busy tdbdat) server-id "shutting-down")
	 ;; (portlogger:open-run-close portlogger:set-port port "released")
	 ;; (exit 1)))
   (cond
    ((not fromdb) (debug:print 3 "WARNING: db:sync-tables called with fromdb missing") -1)
    ((not todb)   (debug:print 3 "WARNING: db:sync-tables called with todb missing") -2)
    ((not (sqlite3:database? (db:dbdat-get-db fromdb)))
     (debug:print 0 "ERROR: db:sync-tables called with fromdb not a database " fromdb) -3)
    ((not (sqlite3:database? (db:dbdat-get-db todb)))
     (debug:print 0 "ERROR: db:sync-tables called with todb not a database " todb) -4)

Modified http-transport.scm from [c0d3a601db] to [0d0f56da13].

390
391
392
393
394
395
396
397
398
399
400

401
402
403
404

405










406
407
408
409





410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
					  (equal? sdat last-sdat)
					  sdat)))))))
         (iface       (car server-info))
         (port        (cadr server-info))
         (last-access 0)
	 (server-timeout (server:get-timeout)))
    (let loop ((count         0)
	       (server-state 'available))


      ;; Use this opportunity to sync the inmemdb to db

      (let ((start-time (current-milliseconds))
	    (sync-time  #f)
	    (rem-time   #f))
	;; inmemdb is a dbstruct

	(if *inmemdb* (db:sync-touched *inmemdb* *run-id* force-sync: #t))










	(set! sync-time  (- (current-milliseconds) start-time))
	(set! rem-time (quotient (- 4000 sync-time) 1000))
	(debug:print 2 "SYNC: time= " sync-time ", rem-time=" rem-time)






      ;;
      ;; set_running after our first pass through and start the db
      ;;
      (if (eq? server-state 'available)
	  (let ((new-server-id (tasks:server-am-i-the-server? (db:delay-if-busy tdbdat) run-id))) ;; try to ensure no double registering of servers
	    (if (equal? new-server-id server-id)
		(begin
		  (tasks:server-set-state! (db:delay-if-busy tdbdat) server-id "dbprep")
		  (thread-sleep! 5) ;; give some margin for queries to complete before switching from file based access to server based access
		  (set! *inmemdb*  (db:setup run-id))
		  (tasks:server-set-state! (db:delay-if-busy tdbdat) server-id "running"))
		(begin ;; gotta exit nicely
		  (tasks:server-set-state! (db:delay-if-busy tdbdat) server-id "collision")
		  (http-transport:server-shutdown server-id port)))))

      (if (and (<= rem-time 4)
	       (> rem-time 0))
	  (thread-sleep! rem-time)
	  (thread-sleep! 4))) ;; fallback for if the math is changed ...
      
      (if (< count 1) ;; 3x3 = 9 secs aprox
	  (loop (+ count 1) 'running))
      
      ;; Check that iface and port have not changed (can happen if server port collides)
      (mutex-lock! *heartbeat-mutex*)
      (set! sdat *server-info*)
      (mutex-unlock! *heartbeat-mutex*)
      
      (if (or (not (equal? sdat (list iface port)))







|
|


>
|
|
|
|
>
|
>
>
>
>
>
>
>
>
>
>
|
|
|
|
>
>
>
>
>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
<
<
<
<
<


|







390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440





441
442
443
444
445
446
447
448
449
450
					  (equal? sdat last-sdat)
					  sdat)))))))
         (iface       (car server-info))
         (port        (cadr server-info))
         (last-access 0)
	 (server-timeout (server:get-timeout)))
    (let loop ((count         0)
	       (server-state 'available)
	       (bad-sync-count 0))

      ;; Use this opportunity to sync the inmemdb to db
      (if *inmemdb* 
	  (let ((start-time (current-milliseconds))
		(sync-time  #f)
		(rem-time   #f))
	    ;; inmemdb is a dbstruct
	    (condition-case
	     (db:sync-touched *inmemdb* *run-id* force-sync: #t)
	     ((sync-failed)(cond
			    ((> bad-sync-count 10) ;; time to give up
			     (http-transport:server-shutdown server-id port))
			    (else ;; (> bad-sync-count 0)  ;; we've had a fail or two, delay and loop
			     (thread-sleep! 5)
			     (loop count server-state (+ bad-sync-count 1)))))
	     ((exn)
	      (debug:print 0 "ERROR: error from sync code other than 'sync-failed. Attempting to gracefully shutdown the server")
	      (tasks:server-delete-record (db:delay-if-busy tdbdat) server-id " http-transport:keep-running crashed")
	      (exit)))
	    (set! sync-time  (- (current-milliseconds) start-time))
	    (set! rem-time (quotient (- 4000 sync-time) 1000))
	    (debug:print 2 "SYNC: time= " sync-time ", rem-time=" rem-time)
	    
	    (if (and (<= rem-time 4)
		     (> rem-time 0))
		(thread-sleep! rem-time)
		(thread-sleep! 4))) ;; fallback for if the math is changed ...

	  ;;
	  ;; no *inmemdb* yet, set running after our first pass through and start the db
	  ;;
	  (if (eq? server-state 'available)
	      (let ((new-server-id (tasks:server-am-i-the-server? (db:delay-if-busy tdbdat) run-id))) ;; try to ensure no double registering of servers
		(if (equal? new-server-id server-id)
		    (begin
		      (tasks:server-set-state! (db:delay-if-busy tdbdat) server-id "dbprep")
		      (thread-sleep! 0.5) ;; give some margin for queries to complete before switching from file based access to server based access
		      (set! *inmemdb*  (db:setup run-id))
		      (tasks:server-set-state! (db:delay-if-busy tdbdat) server-id "running"))
		    (begin ;; gotta exit nicely
		      (tasks:server-set-state! (db:delay-if-busy tdbdat) server-id "collision")
		      (http-transport:server-shutdown server-id port))))))





      
      (if (< count 1) ;; 3x3 = 9 secs aprox
	  (loop (+ count 1) 'running bad-sync-count))
      
      ;; Check that iface and port have not changed (can happen if server port collides)
      (mutex-lock! *heartbeat-mutex*)
      (set! sdat *server-info*)
      (mutex-unlock! *heartbeat-mutex*)
      
      (if (or (not (equal? sdat (list iface port)))
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
	    ;;
	    ;; Consider implementing some smarts here to re-insert the record or kill self is
	    ;; the db indicates so
	    ;;
	    ;; (if (tasks:server-am-i-the-server? tdb run-id)
	    ;;     (tasks:server-set-state! tdb server-id "running"))
	    ;;
	    (loop 0 server-state))
	  (http-transport:server-shutdown server-id port)))))

(define (http-transport:server-shutdown server-id port)
  (let ((tdbdat (tasks:open-db)))
    (debug:print-info 0 "Starting to shutdown the server.")
    ;; need to delete only *my* server entry (future use)
    (set! *time-to-exit* #t)







|







473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
	    ;;
	    ;; Consider implementing some smarts here to re-insert the record or kill self is
	    ;; the db indicates so
	    ;;
	    ;; (if (tasks:server-am-i-the-server? tdb run-id)
	    ;;     (tasks:server-set-state! tdb server-id "running"))
	    ;;
	    (loop 0 server-state bad-sync-count))
	  (http-transport:server-shutdown server-id port)))))

(define (http-transport:server-shutdown server-id port)
  (let ((tdbdat (tasks:open-db)))
    (debug:print-info 0 "Starting to shutdown the server.")
    ;; need to delete only *my* server entry (future use)
    (set! *time-to-exit* #t)

Modified launch.scm from [111f911e42] to [6835480f8d].

586
587
588
589
590
591
592





593
594
595
596
597
598
599
600
    (if (not (file-exists? linktree))
	(begin
	  (debug:print 0 "WARNING: linktree did not exist! Creating it now at " linktree)
	  (create-directory linktree #t))) ;; (system (conc "mkdir -p " linktree))))
    ;; create the directory for the tests dir links, this is needed no matter what...
    (if (and (not (directory-exists? lnkbase))
	     (not (file-exists? lnkbase)))





	(create-directory lnkbase #t))
    
    ;; update the toptest record with its location rundir, cache the path
    ;; This wass highly inefficient, one db write for every subtest, potentially
    ;; thousands of unnecessary updates, cache the fact it was set and don't set it 
    ;; again. 

    ;; Now create the link from the test path to the link tree, however







>
>
>
>
>
|







586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
    (if (not (file-exists? linktree))
	(begin
	  (debug:print 0 "WARNING: linktree did not exist! Creating it now at " linktree)
	  (create-directory linktree #t))) ;; (system (conc "mkdir -p " linktree))))
    ;; create the directory for the tests dir links, this is needed no matter what...
    (if (and (not (directory-exists? lnkbase))
	     (not (file-exists? lnkbase)))
	(handle-exceptions
	 exn
	 (begin
	   (debug:print "ERROR: Problem creating linktree base at " lnkbase)
	   (print-error-message exn (current-error-port)))
	 (create-directory lnkbase #t)))
    
    ;; update the toptest record with its location rundir, cache the path
    ;; This wass highly inefficient, one db write for every subtest, potentially
    ;; thousands of unnecessary updates, cache the fact it was set and don't set it 
    ;; again. 

    ;; Now create the link from the test path to the link tree, however

Modified runs.scm from [411a2b6b6a] to [ad2ef66ee2].

1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647

1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
     ((not target)
      (debug:print 0 "ERROR: Missing required parameter for " switchname ", you must specify the target with -target")
      (exit 3))
     ((not runname)
      (debug:print 0 "ERROR: Missing required parameter for " switchname ", you must specify the run name with -runname runname")
      (exit 3))
     (else
      (let ((db   #f)
	    (keys #f))
	(if (not (launch:setup-for-run))
	    (begin 
	      (debug:print 0 "Failed to setup, exiting")
	      (exit 1)))
	;; (if (args:get-arg "-server")
	;;     (cdb:remote-run server:start db (args:get-arg "-server")))
	(set! keys (keys:config-get-fields *configdat*))
	;; have enough to process -target or -reqtarg here
	(if (args:get-arg "-reqtarg")
	    (let* ((runconfigf (conc  *toppath* "/runconfigs.config")) ;; DO NOT EVALUATE ALL 
		   (runconfig  (read-config runconfigf #f #t environ-patt: #f)))
	      (if (hash-table-ref/default runconfig (args:get-arg "-reqtarg") #f)
		  (keys:target-set-args keys (args:get-arg "-reqtarg") args:arg-hash)
		    
		  (begin
		    (debug:print 0 "ERROR: [" (args:get-arg "-reqtarg") "] not found in " runconfigf)
		    (if db (sqlite3:finalize! db))
		    (exit 1))))

	    (if (args:get-arg "-target")
		(keys:target-set-args keys (args:get-arg "-target" args:arg-hash) args:arg-hash)))
	(if (not (car *configinfo*))
	    (begin
	      (debug:print 0 "ERROR: Attempted to " action-desc " but run area config file not found")
	      (exit 1))
	    ;; Extract out stuff needed in most or many calls
	    ;; here then call proc
	    (let* ((keyvals    (keys:target->keyval keys target)))
	      (proc target runname keys keyvals)))
	(if db (sqlite3:finalize! db))
	(set! *didsomething* #t))))))

;;======================================================================
;; Lock/unlock runs
;;======================================================================

(define (runs:handle-locking target keys runname lock unlock user)







|

















|
|
>










|







1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
     ((not target)
      (debug:print 0 "ERROR: Missing required parameter for " switchname ", you must specify the target with -target")
      (exit 3))
     ((not runname)
      (debug:print 0 "ERROR: Missing required parameter for " switchname ", you must specify the run name with -runname runname")
      (exit 3))
     (else
      (let (;; (db   #f)
	    (keys #f))
	(if (not (launch:setup-for-run))
	    (begin 
	      (debug:print 0 "Failed to setup, exiting")
	      (exit 1)))
	;; (if (args:get-arg "-server")
	;;     (cdb:remote-run server:start db (args:get-arg "-server")))
	(set! keys (keys:config-get-fields *configdat*))
	;; have enough to process -target or -reqtarg here
	(if (args:get-arg "-reqtarg")
	    (let* ((runconfigf (conc  *toppath* "/runconfigs.config")) ;; DO NOT EVALUATE ALL 
		   (runconfig  (read-config runconfigf #f #t environ-patt: #f)))
	      (if (hash-table-ref/default runconfig (args:get-arg "-reqtarg") #f)
		  (keys:target-set-args keys (args:get-arg "-reqtarg") args:arg-hash)
		    
		  (begin
		    (debug:print 0 "ERROR: [" (args:get-arg "-reqtarg") "] not found in " runconfigf)
		    ;; (if db (sqlite3:finalize! db))
		    (exit 1)
		    )))
	    (if (args:get-arg "-target")
		(keys:target-set-args keys (args:get-arg "-target" args:arg-hash) args:arg-hash)))
	(if (not (car *configinfo*))
	    (begin
	      (debug:print 0 "ERROR: Attempted to " action-desc " but run area config file not found")
	      (exit 1))
	    ;; Extract out stuff needed in most or many calls
	    ;; here then call proc
	    (let* ((keyvals    (keys:target->keyval keys target)))
	      (proc target runname keys keyvals)))
	;; (if db (sqlite3:finalize! db))
	(set! *didsomething* #t))))))

;;======================================================================
;; Lock/unlock runs
;;======================================================================

(define (runs:handle-locking target keys runname lock unlock user)