| ︙ | | | ︙ | |
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
|
(declare (unit common))
(declare (uses commonmod))
(import commonmod)
(include "common_records.scm")
;; (require-library margs)
;; (include "margs.scm")
;; (define old-exit exit)
;;
;; (define (exit . code)
;; (if (null? code)
;; (old-exit)
;; (old-exit code)))
(define (stop-the-train)
(thread-start! (make-thread (lambda ()
(let loop ()
(if (and *toppath*
(file-exists? (conc *toppath*"/stop-the-train")))
(begin
(debug:print 0 *default-log-port* "ERROR: found file "*toppath*"/stop-the-train, exiting immediately")
(exit 1)))
(thread-sleep! 5)
(loop))))))
;; execute thunk, return value. If exception thrown, trap exception, return #f, and emit nonfatal condition note to *default-log-port* .
;; arguments - thunk, message
(define (common:fail-safe thunk warning-message-on-exception)
|
|
|
<
|
<
<
<
<
<
<
>
>
>
|
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
|
(declare (unit common))
(declare (uses commonmod))
(import commonmod)
(include "common_records.scm")
(define (remove-files filespec)
(let ((files (glob filespec)))
(for-each delete-file* files)))
(define (stop-the-train)
(thread-start! (make-thread (lambda ()
(let loop ()
(if (and *toppath*
(file-exists? (conc *toppath*"/stop-the-train")))
(begin
(debug:print 0 *default-log-port* "ERROR: found file "*toppath*"/stop-the-train, exiting immediately")
(remove-files (conc *toppath* "/logs/server*"))
(remove-files (conc *toppath* "/.servinfo/*"))
(remove-files (conc *toppath* "/.mtdb/*lock"))
(exit 1)))
(thread-sleep! 5)
(loop))))))
;; execute thunk, return value. If exception thrown, trap exception, return #f, and emit nonfatal condition note to *default-log-port* .
;; arguments - thunk, message
(define (common:fail-safe thunk warning-message-on-exception)
|
| ︙ | | | ︙ | |
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
|
(define *pkts-info* (make-hash-table)) ;; store stuff like the last parent here
(define *configinfo* #f) ;; raw results from setup, includes toppath and table from megatest.config
(define *runconfigdat* #f) ;; run configs data
(define *configdat* #f) ;; megatest.config data
(define *configstatus* #f) ;; status of data; 'fulldata : all processing done, #f : no data yet, 'partialdata : partial read done
(define *toppath* #f)
(define *already-seen-runconfig-info* #f)
(define *test-meta-updated* (make-hash-table))
(define *globalexitstatus* 0) ;; attempt to work around possible thread issues
(define *passnum* 0) ;; when running track calls to run-tests or similar
;; (define *alt-log-file* #f) ;; used by -log
;; (define *common:denoise* (make-hash-table)) ;; for low noise printing
(define *default-log-port* (current-error-port))
(define *time-zero* (current-seconds)) ;; for the watchdog
(define *default-area-tag* "local")
;; DATABASE
;; (define *dbstruct-dbs* #f) ;; used to cache the dbstruct in db:setup. Goal is to remove this.
;; db stats
(define *db-stats* (make-hash-table)) ;; hash of vectors < count duration-total >
(define *db-stats-mutex* (make-mutex))
;; db access
(define *db-last-access* (current-seconds)) ;; last db access, used in server
;; (define *db-write-access* #t)
;; db sync
;; (define *db-last-sync* 0) ;; last time the sync to megatest.db happened
(define *db-sync-in-progress* #f) ;; if there is a sync in progress do not try to start another
;; (define *db-multi-sync-mutex* (make-mutex)) ;; protect access to *db-sync-in-progress*, *db-last-sync*
;; task db
(define *task-db* #f) ;; (vector db path-to-db)
(define *db-access-allowed* #t) ;; flag to allow access
;; (define *db-access-mutex* (make-mutex)) ;; moved to dbfile
(define *db-transaction-mutex* (make-mutex))
|
>
|
|
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
|
(define *pkts-info* (make-hash-table)) ;; store stuff like the last parent here
(define *configinfo* #f) ;; raw results from setup, includes toppath and table from megatest.config
(define *runconfigdat* #f) ;; run configs data
(define *configdat* #f) ;; megatest.config data
(define *configstatus* #f) ;; status of data; 'fulldata : all processing done, #f : no data yet, 'partialdata : partial read done
(define *toppath* #f)
(define *dbdir* ".mtdb")
(define *already-seen-runconfig-info* #f)
(define *test-meta-updated* (make-hash-table))
(define *globalexitstatus* 0) ;; attempt to work around possible thread issues
(define *passnum* 0) ;; when running track calls to run-tests or similar
;; (define *alt-log-file* #f) ;; used by -log
;; (define *common:denoise* (make-hash-table)) ;; for low noise printing
(define *default-log-port* (current-error-port))
(define *time-zero* (current-seconds)) ;; for the watchdog
(define *default-area-tag* "local")
;; DATABASE
;; (define *dbstruct-dbs* #f) ;; used to cache the dbstruct in db:setup. Goal is to remove this.
;; db stats
(define *db-stats* (make-hash-table)) ;; hash of vectors < count duration-total >
(define *db-stats-mutex* (make-mutex))
;; db access
(define *db-last-access* (current-seconds)) ;; last db access, used in server
;; (define *db-write-access* #t)
;; db sync
;; (define *db-last-sync* 0) ;; last time the sync to nfs db happened
(define *db-sync-in-progress* #f) ;; if there is a sync in progress do not try to start another
;; (define *db-multi-sync-mutex* (make-mutex)) ;; protect access to *db-sync-in-progress*, *db-last-sync*
;; task db
(define *task-db* #f) ;; (vector db path-to-db)
(define *db-access-allowed* #t) ;; flag to allow access
;; (define *db-access-mutex* (make-mutex)) ;; moved to dbfile
(define *db-transaction-mutex* (make-mutex))
|
| ︙ | | | ︙ | |
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
|
(debug:print-info 2 *default-log-port* "unable to rotate log " file ", probably handled by another process, this is safe to ignore. exn=" exn)
(debug:print 2 *default-log-port* " message: " ((condition-property-accessor 'exn 'message) exn))
;; (print-call-chain (current-error-port)) ;;
)
(let* ((fullname (conc "logs/" file))
(mod-time (file-modification-time fullname))
(file-age (- (current-seconds) mod-time))
(file-old (> file-age (* 48 60 60)))
(file-big (> (file-size fullname) 200000)))
(hash-table-set! all-files file mod-time)
(if (or (and (string-match "^.*.log" file)
|
|
>
>
>
|
>
>
|
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
|
(debug:print-info 2 *default-log-port* "unable to rotate log " file ", probably handled by another process, this is safe to ignore. exn=" exn)
(debug:print 2 *default-log-port* " message: " ((condition-property-accessor 'exn 'message) exn))
;; (print-call-chain (current-error-port)) ;;
)
(let* ((fullname (conc "logs/" file))
(mod-time (file-modification-time fullname))
(file-age (- (current-seconds) mod-time))
(file-old (> file-age (* 48 60 60))) ;; over 48 hours
(file-big (> (file-size fullname) 200000)))
(hash-table-set! all-files file mod-time)
(if (or
;; gzip:
;; any old and big log files: (server logs, runlogs, update_ext_specs, etc.
(and (string-match "^.*.log" file)
file-old
file-big)
;; old server log files:
(and (string-match "^server-.*.log" file)
file-old))
(let ((gzfile (conc fullname ".gz")))
(if (common:file-exists? gzfile)
(begin
(debug:print-info 0 *default-log-port* "removing " gzfile)
(delete-file* gzfile)
(hash-table-delete! all-files gzfile) ;; needed?
))
(debug:print-info 0 *default-log-port* "compressing " file)
(system (conc "gzip " fullname))
(inc-stat "gzipped")
(hash-table-set! all-files (conc file ".gz") file-age) ;; add the .gz file and remove the base file
(hash-table-delete! all-files file)
)
;; delete other files over expiration date:
(if (and (> file-age (* (string->number (or (configf:lookup *configdat* "setup" "log-expire-days") "30")) 24 3600))
(file-exists? fullname)) ;; just in case it was gzipped - will get it next time
(handle-exceptions
exn
#f
(if (directory? fullname)
(begin
|
| ︙ | | | ︙ | |
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
|
(if (> num-logs max-allowed) ;; because NFS => don't let number of logs exceed 300
(let ((files (take (sort (hash-table-keys all-files)
(lambda (a b)
(< (hash-table-ref all-files a)(hash-table-ref all-files b))))
(- num-logs max-allowed))))
(for-each
(lambda (file)
(let* ((fullname (conc "logs/" file)))
(if (directory? fullname)
(debug:print-info 0 *default-log-port* fullname " in logs directory is a directory! Cannot rotate it, it is best to not put subdirectories in the logs dir.")
(handle-exceptions
exn
(debug:print-error 0 *default-log-port* "failed to remove " fullname ", exn=" exn)
(delete-file* fullname)))))
;;======================================================================
;; Force a megatest cleanup-db if version is changed and skip-version-check not specified
;; Do NOT check if not on homehost!
;;
(define (common:exit-on-version-changed)
(if (common:on-homehost?)
(if (common:api-changed?)
(let* ((mtconf (conc (get-environment-variable "MT_RUN_AREA_HOME") "/megatest.config"))
(dbfile (conc (get-environment-variable "MT_RUN_AREA_HOME") "/megatest.db"))
(read-only (not (file-write-access? dbfile)))
(dbstruct (db:setup #t))) ;; (db:setup-db *dbstruct-dbs* *toppath* #f))) ;; #t)))
(debug:print 0 *default-log-port*
"WARNING: Version mismatch!\n"
" expected: " (common:version-signature) "\n"
" got: " (common:get-last-run-version))
(cond
((get-environment-variable "MT_SKIP_DB_MIGRATE") #t)
((and (common:file-exists? mtconf) (common:file-exists? dbfile) (not read-only)
(eq? (current-user-id)(file-owner mtconf))) ;; safe to run -cleanup-db
(debug:print 0 *default-log-port* " I see you are the owner of megatest.config, attempting to cleanup and reset to new version")
(handle-exceptions
exn
(begin
(debug:print 0 *default-log-port* "Failed to switch versions. exn=" exn)
(debug:print 0 *default-log-port* " message: " ((condition-property-accessor 'exn 'message) exn))
(print-call-chain (current-error-port))
(exit 1))
(common:cleanup-db dbstruct)))
((not (common:file-exists? mtconf))
(debug:print 0 *default-log-port* " megatest.config does not exist in this area. Cannot proceed with megatest version migration.")
(exit 1))
((not (common:file-exists? dbfile))
(debug:print 0 *default-log-port* " megatest.db does not exist in this area. Cannot proceed with megatest version migration.")
(exit 1))
((not (eq? (current-user-id)(file-owner mtconf)))
(debug:print 0 *default-log-port* " You do not own megatest.db in this area. Cannot proceed with megatest version migration.")
(exit 1))
(read-only
(debug:print 0 *default-log-port* " You have read-only access to this area. Cannot proceed with megatest version migration.")
(exit 1))
(else
(debug:print 0 *default-log-port* " to switch versions you can run: \"megatest -cleanup-db\"")
(exit 1)))))))
|
|
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
|
>
>
>
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
|
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
|
(if (> num-logs max-allowed) ;; because NFS => don't let number of logs exceed 300
(let ((files (take (sort (hash-table-keys all-files)
(lambda (a b)
(< (hash-table-ref all-files a)(hash-table-ref all-files b))))
(- num-logs max-allowed))))
(for-each
(lambda (file)
(let* ((fullname (conc "logs/" file))
(is-alive 0))
;; Don't delete it if it's the log file of a running server.
(if (string-match "server-\\d+-[a-zA-Z0-9]+\\.log" file)
(let* ((parts (string-split file "-."))
(pid (cadr parts)) ; Second part is the PID
(server-machine (caddr parts)) ; Third part is the server machine
(local (string=? (get-host-name) server-machine))
(test-cmd (conc "test -d /proc/" pid)))
(if local
(set! is-alive (not (system test-cmd)))
(set! is-alive (not (system (conc "ssh " server-machine " test -d /proc/" pid))))
)
)
)
(if (directory? fullname)
(debug:print-info 0 *default-log-port* fullname " in logs directory is a directory! Cannot rotate it, it is best to not put subdirectories in the logs dir.")
(handle-exceptions
exn
(debug:print-error 0 *default-log-port* "failed to remove " fullname ", exn=" exn)
(if (not is-alive)
(delete-file* fullname)
(debug:print-info 0 *default-log-port* "Not deleting log file " file " since its server is still alive")
)
))))
files)
(debug:print-info 0 *default-log-port* "Deleted " (length files) " files from logs, keeping " max-allowed " files."))))))
;;======================================================================
;; (begin
;; (debug:print 0 *default-log-port* "ERROR: cannot migrate version unless on homehost. Exiting.")
;; (exit 1))))
;;======================================================================
;; S P A R S E A R R A Y S
|
| ︙ | | | ︙ | |
704
705
706
707
708
709
710
711
712
713
714
715
716
717
|
(if dat
dat
""))))
(define (common:alist-ref/default key alist default)
(or (alist-ref key alist) default))
(define (common:low-noise-print waitval . keys)
(let* ((key (string-intersperse (map conc keys) "-" ))
(lasttime (hash-table-ref/default *common:denoise* key 0))
(currtime (current-seconds)))
(if (> (- currtime lasttime) waitval)
(begin
(hash-table-set! *common:denoise* key currtime)
|
>
>
>
>
>
>
|
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
|
(if dat
dat
""))))
(define (common:alist-ref/default key alist default)
(or (alist-ref key alist) default))
;; The `common:low-noise-print` function is a utility that can be used to throttle the
;; frequency of certain operations. It does this by tracking the last time an operation was
;; performed and only allowing it again after a specified interval (`waitval`). This can be useful
;; for reducing noise in logs or limiting the rate of user notifications, among other use cases.
(define (common:low-noise-print waitval . keys)
(let* ((key (string-intersperse (map conc keys) "-" ))
(lasttime (hash-table-ref/default *common:denoise* key 0))
(currtime (current-seconds)))
(if (> (- currtime lasttime) waitval)
(begin
(hash-table-set! *common:denoise* key currtime)
|
| ︙ | | | ︙ | |
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
|
(string-translate *toppath* "/" "."))
(conc "/tmp/" (current-process-id) ;; just in case we have an issue with the dir by own user name
"/megatest_localdb/"
tsname
(string-translate *toppath* "/" "."))
))))
(set! *db-cache-path* dbpath)
;; ensure megatest area has .megatest
(let ((dbarea (conc *toppath* "/.megatest")))
(if (not (file-exists? dbarea))
(create-directory dbarea)))
;; ensure tmp area has .megatest
(let ((dbarea (conc dbpath "/.megatest")))
(if (not (file-exists? dbarea))
(create-directory dbarea)))
dbpath))
#f)))
(define (common:get-area-path-signature)
(message-digest-string (md5-primitive) *toppath*))
|
|
|
|
|
|
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
|
(string-translate *toppath* "/" "."))
(conc "/tmp/" (current-process-id) ;; just in case we have an issue with the dir by own user name
"/megatest_localdb/"
tsname
(string-translate *toppath* "/" "."))
))))
(set! *db-cache-path* dbpath)
;; ensure megatest area has dbdir
(let ((dbarea (conc *toppath* "/" *dbdir*)))
(if (not (file-exists? dbarea))
(create-directory dbarea)))
;; ensure tmp area has dbdir
(let ((dbarea (conc dbpath "/" *dbdir*)))
(if (not (file-exists? dbarea))
(create-directory dbarea)))
dbpath))
#f)))
(define (common:get-area-path-signature)
(message-digest-string (md5-primitive) *toppath*))
|
| ︙ | | | ︙ | |