| ︙ | | | ︙ | |
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
|
matchable regex posix (srfi 18) extras ;; tcp
(prefix nanomsg nmsg:)
(prefix sqlite3 sqlite3:)
pkts (prefix dbi dbi:)
)
(declare (unit common))
;; (declare (uses commonmod))
;; (import commonmod)
|
|
|
|
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
|
matchable regex posix (srfi 18) extras ;; tcp
(prefix nanomsg nmsg:)
(prefix sqlite3 sqlite3:)
pkts (prefix dbi dbi:)
)
(declare (unit common))
(declare (uses commonmod))
(import commonmod)
(include "common_records.scm")
;; (require-library margs)
;; (include "margs.scm")
|
| ︙ | | | ︙ | |
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
181
182
183
184
185
|
(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-db* #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))
(define *db-transaction-mutex* (make-mutex))
(define *db-cache-path* #f)
(define *db-with-db-mutex* (make-mutex))
(define *db-api-call-time* (make-hash-table)) ;; hash of command => (list of times)
;; no sync db
(define *no-sync-db* #f)
;; SERVER
(define *my-client-signature* #f)
(define *transport-type* 'http) ;; override with [server] transport http|rpc|nmsg
(define *runremote* #f) ;; if set up for server communication this will hold <host port>
;; (define *max-cache-size* 0)
(define *logged-in-clients* (make-hash-table))
(define *server-id* #f)
(define *server-info* #f) ;; good candidate for easily convert to non-global
(define *time-to-exit* #f)
(define *server-run* #t)
(define *run-id* #f)
(define *server-kind-run* (make-hash-table))
(define *home-host* #f)
;; (define *total-non-write-delay* 0)
(define *heartbeat-mutex* (make-mutex))
(define *api-process-request-count* 0)
(define *max-api-process-requests* 0)
(define *server-overloaded* #f)
;; client
(define *rmt-mutex* (make-mutex)) ;; remote access calls mutex
;; RPC transport
(define *rpc:listener* #f)
|
|
|
|
|
|
|
|
|
|
|
|
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
181
182
183
184
185
|
(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))
(define *db-cache-path* #f)
;; (define *db-with-db-mutex* (make-mutex))
(define *db-api-call-time* (make-hash-table)) ;; hash of command => (list of times)
;; no sync db
;; (define *no-sync-db* #f) ;; moved to dbfile
;; SERVER
(define *my-client-signature* #f)
(define *transport-type* 'http) ;; override with [server] transport http|rpc|nmsg
(define *runremote* #f) ;; if set up for server communication this will hold <host port>
;; (define *max-cache-size* 0)
(define *logged-in-clients* (make-hash-table))
(define *server-id* #f)
(define *server-info* #f) ;; good candidate for easily convert to non-global
(define *time-to-exit* #f)
(define *server-run* #t)
(define *run-id* #f)
(define *server-kind-run* (make-hash-table))
(define *home-host* #f)
;; (define *total-non-write-delay* 0)
(define *heartbeat-mutex* (make-mutex))
;; (define *api-process-request-count* 0)
;; (define *max-api-process-requests* 0)
(define *server-overloaded* #f)
;; client
(define *rmt-mutex* (make-mutex)) ;; remote access calls mutex
;; RPC transport
(define *rpc:listener* #f)
|
| ︙ | | | ︙ | |
308
309
310
311
312
313
314
315
316
317
318
319
320
321
|
(defstruct remote
(hh-dat (common:get-homehost)) ;; homehost record ( addr . hhflag )
(server-url #f) ;; (server:check-if-running *toppath*) #f))
(server-id #f)
(server-info (if *toppath* (server:check-if-running *toppath*) #f))
(last-server-check 0) ;; last time we checked to see if the server was alive
(conndat #f)
(transport *transport-type*)
(server-timeout (server:expiration-timeout))
(force-server #f)
(ro-mode #f)
(ro-mode-checked #f)) ;; flag that indicates we have checked for ro-mode
|
>
|
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
|
(defstruct remote
(hh-dat (common:get-homehost)) ;; homehost record ( addr . hhflag )
(server-url #f) ;; (server:check-if-running *toppath*) #f))
(server-id #f)
(server-info (if *toppath* (server:check-if-running *toppath*) #f))
(last-server-check 0) ;; last time we checked to see if the server was alive
(connect-time (current-seconds))
(conndat #f)
(transport *transport-type*)
(server-timeout (server:expiration-timeout))
(force-server #f)
(ro-mode #f)
(ro-mode-checked #f)) ;; flag that indicates we have checked for ro-mode
|
| ︙ | | | ︙ | |
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
|
;; Move me elsewhere ...
;; RADT => Why do we meed the version check here, this is called only if version misma
;;
(define (common:cleanup-db dbstruct #!key (full #f))
(apply db:multi-db-sync
dbstruct
'schema
;; 'new2old
;; 'old2new
;; (if full
'(dejunk)
;; '())
)
|
<
<
<
|
<
|
|
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
|
;; Move me elsewhere ...
;; RADT => Why do we meed the version check here, this is called only if version misma
;;
(define (common:cleanup-db dbstruct #!key (full #f))
(apply db:multi-db-sync
dbstruct
'schema
'killservers
'adj-target
'new2old
'(dejunk)
)
(if (common:api-changed?)
(common:set-last-run-version)))
(define (common:snapshot-file filepath #!key (subdir ".") )
(if (file-exists? filepath)
(let* ((age-sec (lambda (file)
(if (file-exists? file)
|
| ︙ | | | ︙ | |
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
|
;; 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)))
(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)
|
|
|
|
|
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
|
;; 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)
|
| ︙ | | | ︙ | |
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
|
(begin
(debug:print-error 0 *default-log-port* "received bad encoded string \"" instr "\", message: " ((condition-property-accessor 'exn 'message) exn) ", exn=" exn)
(print-call-chain (current-error-port))
#f)
(read (open-input-string (base64:base64-decode instr))))
(read (open-input-string (z3:decode-buffer (base64:base64-decode instr))))))
;; dot-locking egg seems not to work, using this for now
;; if lock is older than expire-time then remove it and try again
;; to get the lock
;;
(define (common:simple-file-lock fname #!key (expire-time 300))
(let ((fmod-time (handle-exceptions
ext
(current-seconds)
(file-modification-time fname))))
(if (common:file-exists? fname)
(if (> (- (current-seconds) fmod-time) expire-time)
(begin
(handle-exceptions exn #f (delete-file* fname))
(common:simple-file-lock fname expire-time: expire-time))
#f)
(let ((key-string (conc (get-host-name) "-" (current-process-id))))
(with-output-to-file fname
(lambda ()
(print key-string)))
(thread-sleep! 0.25)
(if (common:file-exists? fname)
(handle-exceptions exn
#f
(with-input-from-file fname
(lambda ()
(equal? key-string (read-line)))))
#f)))))
(define (common:simple-file-lock-and-wait fname #!key (expire-time 300))
(let ((end-time (+ expire-time (current-seconds))))
(let loop ((got-lock (common:simple-file-lock fname expire-time: expire-time)))
(if got-lock
#t
(if (> end-time (current-seconds))
(begin
(thread-sleep! 3)
(loop (common:simple-file-lock fname expire-time: expire-time)))
#f)))))
(define (common:simple-file-release-lock fname)
(handle-exceptions
exn
#f ;; I don't really care why this failed (at least for now)
(delete-file* fname)))
|
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
|
719
720
721
722
723
724
725
726
727
728
729
730
731
732
|
(begin
(debug:print-error 0 *default-log-port* "received bad encoded string \"" instr "\", message: " ((condition-property-accessor 'exn 'message) exn) ", exn=" exn)
(print-call-chain (current-error-port))
#f)
(read (open-input-string (base64:base64-decode instr))))
(read (open-input-string (z3:decode-buffer (base64:base64-decode instr))))))
;;======================================================================
;; S T A T E S A N D S T A T U S E S
;;======================================================================
;; BBnote: *common:std-states* - dashboard filter control and test control state buttons defined here; used in set-fields-panel and dboard:make-controls
(define *common:std-states* ;; for toggle buttons in dashboard
|
| ︙ | | | ︙ | |
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
|
(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)
dbpath))
#f)))
(define (common:get-area-path-signature)
(message-digest-string (md5-primitive) *toppath*))
(define (common:get-signature str)
(message-digest-string (md5-primitive) str))
;;======================================================================
;; E X I T H A N D L I N G
;;======================================================================
(define (common:run-sync?)
(and (common:on-homehost?)
(args:get-arg "-server")))
(define (common:human-time)
(time->string (seconds->local-time (current-seconds)) "%Y-%m-%d %H:%M:%S"))
;;======================================================================
;; currently the primary job of the watchdog is to run the sync back to megatest.db from the db in /tmp
;; if we are on the homehost and we are a server (by definition we are on the homehost if we are a server)
;;
(define (common:readonly-watchdog dbstruct)
(thread-sleep! 0.05) ;; delay for startup
(debug:print-info 13 *default-log-port* "common:readonly-watchdog entered.")
;; sync megatest.db to /tmp/.../megatst.db
(let* ((sync-cool-off-duration 3)
(golden-mtdb (dbr:dbstruct-mtdb dbstruct))
(golden-mtpath (db:dbdat-get-path golden-mtdb))
(tmp-mtdb (dbr:dbstruct-tmpdb dbstruct))
(tmp-mtpath (db:dbdat-get-path tmp-mtdb)))
(debug:print-info 0 *default-log-port* "Read-only periodic sync thread started.")
(let loop ((last-sync-time 0))
(debug:print-info 13 *default-log-port* "loop top tmp-mtpath="tmp-mtpath" golden-mtpath="golden-mtpath)
(let* ((duration-since-last-sync (- (current-seconds) last-sync-time)))
(debug:print-info 13 *default-log-port* "duration-since-last-sync="duration-since-last-sync)
(if (and (not *time-to-exit*)
(< duration-since-last-sync sync-cool-off-duration))
(thread-sleep! (- sync-cool-off-duration duration-since-last-sync)))
(if (not *time-to-exit*)
(let ((golden-mtdb-mtime (file-modification-time golden-mtpath))
(tmp-mtdb-mtime (file-modification-time tmp-mtpath)))
(if (> golden-mtdb-mtime tmp-mtdb-mtime)
(if (< golden-mtdb-mtime (- (current-seconds) 3)) ;; file has NOT been touched in past three seconds, this way multiple servers won't fight to sync back
(let ((res (db:multi-db-sync dbstruct 'old2new)))
(debug:print-info 13 *default-log-port* "rosync called, " res " records transferred."))))
(loop (current-seconds)))
#t)))
(debug:print-info 0 *default-log-port* "Exiting readonly-watchdog timer, *time-to-exit* = " *time-to-exit*" pid="(current-process-id)" mtpath="golden-mtpath)))
;;======================================================================
;; TODO: for multiple areas, we will have multiple watchdogs; and multiple threads to manage
(define (common:watchdog)
(debug:print-info 13 *default-log-port* "common:watchdog entered.")
(if (launch:setup)
(if (common:on-homehost?)
(let ((dbstruct (db:setup #t)))
(debug:print-info 13 *default-log-port* "after db:setup with dbstruct=" dbstruct)
(cond
((dbr:dbstruct-read-only dbstruct)
(debug:print-info 13 *default-log-port* "loading read-only watchdog")
(common:readonly-watchdog dbstruct))
(else
(debug:print-info 13 *default-log-port* "loading writable-watchdog.")
(let* ((syncer (or (configf:lookup *configdat* "server" "sync-method") "brute-force-sync")))
(cond
((equal? syncer "brute-force-sync")
(server:writable-watchdog-bruteforce dbstruct))
((equal? syncer "delta-sync")
(server:writable-watchdog-deltasync dbstruct))
(else
(debug:print-error 0 *default-log-port* "Unknown server/sync-method specified ("syncer") - valid values are brute-force-sync and delta-sync.")
(exit 1)))
;;(debug:print 1 *default-log-port* "INFO: ["(common:human-time)"] Syncer started (method="syncer")")
)))
(debug:print-info 13 *default-log-port* "watchdog done."))
(debug:print-info 13 *default-log-port* "no need for watchdog on non-homehost"))))
(define (std-exit-procedure)
;;(common:telemetry-log-close)
(on-exit (lambda () 0))
;;(debug:print-info 13 *default-log-port* "std-exit-procedure called; *time-to-exit*="*time-to-exit*)
(let ((no-hurry (if *time-to-exit* ;; hurry up
#f
(begin
(set! *time-to-exit* #t)
#t))))
(debug:print-info 4 *default-log-port* "starting exit process, finalizing databases.")
(if (and no-hurry (debug:debug-mode 18))
(rmt:print-db-stats))
(let ((th1 (make-thread (lambda () ;; thread for cleaning up, give it five seconds
(if *dbstruct-db* (db:close-all *dbstruct-db*)) ;; one second allocated
(if *task-db*
(let ((db (cdr *task-db*)))
(if (sqlite3:database? db)
(begin
(sqlite3:interrupt! db)
(sqlite3:finalize! db #t)
;; (vector-set! *task-db* 0 #f)
(set! *task-db* #f)))))
(http-client#close-all-connections!)
;; (if (and *runremote*
;; (remote-conndat *runremote*))
;; (begin
;; (http-client#close-all-connections!))) ;; for http-client
(if (not (eq? *default-log-port* (current-error-port)))
(close-output-port *default-log-port*))
(set! *default-log-port* (current-error-port))) "Cleanup db exit thread"))
(th2 (make-thread (lambda ()
(debug:print 4 *default-log-port* "Attempting clean exit. Please be patient and wait a few seconds...")
(if no-hurry
(begin
(thread-sleep! 5)) ;; give the clean up few seconds to do it's stuff
(begin
(thread-sleep! 2)))
(debug:print 4 *default-log-port* " ... done")
)
"clean exit")))
(thread-start! th1)
(thread-start! th2)
(thread-join! th1)
)
)
0)
|
>
>
>
>
>
>
>
>
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
|
930
931
932
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
|
(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*))
(define (common:get-signature str)
(message-digest-string (md5-primitive) str))
;;======================================================================
;; E X I T H A N D L I N G
;;======================================================================
(define (common:run-sync?)
(and (common:on-homehost?)
(args:get-arg "-server")))
(define (common:human-time)
(time->string (seconds->local-time (current-seconds)) "%Y-%m-%d %H:%M:%S"))
(define (std-signal-handler signum)
;; (signal-mask! signum)
(set! *time-to-exit* #t)
;;(debug:print-info 13 *default-log-port* "got signal "signum)
(debug:print-error 0 *default-log-port* "Received signal " signum " aaa exiting promptly")
;; (std-exit-procedure) ;; shouldn't need this since we are exiting and it will be called anyway
|
| ︙ | | | ︙ | |