Overview
| Comment: | merged with latest v1.63 |
|---|---|
| Downloads: | Tarball | ZIP archive | SQL archive |
| Timelines: | family | ancestors | descendants | both | v1.63-configdbsync |
| Files: | files | file ages | folders |
| SHA1: |
7e694854593a05cefa2d52deaaca276d |
| User & Date: | srehman on 2017-01-03 14:03:13 |
| Other Links: | branch diff | manifest | tags |
Context
|
2017-01-09
| ||
| 12:04 | updated to latest v1.63, WIP check-in: 8e63364af0 user: srehman tags: v1.63-configdbsync | |
|
2017-01-03
| ||
| 14:03 | merged with latest v1.63 check-in: 7e69485459 user: srehman tags: v1.63-configdbsync | |
| 10:50 | Commented out a call to BB> that seemed to have trouble ... check-in: e26b0892dd user: mrwellan tags: v1.63 | |
|
2016-12-23
| ||
| 12:28 | added ability to sync delta changes with configdb check-in: 57bb8e5d33 user: srehman tags: v1.63-configdbsync | |
Changes
Modified api.scm from [19e6c44e9d] to [1debddd502].
| ︙ | ︙ | |||
60 61 62 63 64 65 66 67 68 69 70 71 72 73 |
testmeta-get-record
have-incompletes?
synchash-get
))
(define api:write-queries
'(
;; SERVERS
start-server
kill-server
;; TESTS
test-set-state-status-by-id
delete-test-records
| > > | 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 |
testmeta-get-record
have-incompletes?
synchash-get
))
(define api:write-queries
'(
get-keys-write ;; dummy "write" query to force server start
;; SERVERS
start-server
kill-server
;; TESTS
test-set-state-status-by-id
delete-test-records
|
| ︙ | ︙ | |||
187 188 189 190 191 192 193 194 195 196 197 198 199 200 | ;;====================================================================== ;; READ ONLY QUERIES ;;====================================================================== ;; KEYS ((get-key-val-pairs) (apply db:get-key-val-pairs dbstruct params)) ((get-keys) (db:get-keys dbstruct)) ((get-key-vals) (apply db:get-key-vals dbstruct params)) ((get-target) (apply db:get-target dbstruct params)) ((get-targets) (db:get-targets dbstruct)) ;; ARCHIVES ((test-get-archive-block-info) (apply db:test-get-archive-block-info dbstruct params)) | > | 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 |
;;======================================================================
;; READ ONLY QUERIES
;;======================================================================
;; KEYS
((get-key-val-pairs) (apply db:get-key-val-pairs dbstruct params))
((get-keys) (db:get-keys dbstruct))
((get-keys-write) (db:get-keys dbstruct)) ;; force a dummy "write" query to force server
((get-key-vals) (apply db:get-key-vals dbstruct params))
((get-target) (apply db:get-target dbstruct params))
((get-targets) (db:get-targets dbstruct))
;; ARCHIVES
((test-get-archive-block-info) (apply db:test-get-archive-block-info dbstruct params))
|
| ︙ | ︙ |
Modified common.scm from [7213eed61f] to [d2df0eaa3f].
| ︙ | ︙ | |||
551 552 553 554 555 556 557 |
;;;; run-ids
;; if #f use *db-local-sync* : or 'local-sync-flags
;; if #t use timestamps : or 'timestamps
(define (common:sync-to-megatest.db dbstruct)
(let ((start-time (current-seconds))
(res (db:multi-db-sync dbstruct 'new2old)))
(let ((sync-time (- (current-seconds) start-time)))
| | | > > > > > > | > > | | > < | > | | 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 |
;;;; run-ids
;; if #f use *db-local-sync* : or 'local-sync-flags
;; if #t use timestamps : or 'timestamps
(define (common:sync-to-megatest.db dbstruct)
(let ((start-time (current-seconds))
(res (db:multi-db-sync dbstruct 'new2old)))
(let ((sync-time (- (current-seconds) start-time)))
(debug:print-info 3 *default-log-port* "Sync of newdb to olddb completed in " sync-time " seconds pid="(current-process-id))
(if (common:low-noise-print 30 "sync new to old")
(debug:print-info 0 *default-log-port* "Sync of newdb to olddb completed in " sync-time " seconds pid="(current-process-id))))
res))
(define *wdnum* 0)
(define *wdnum*mutex (make-mutex))
;; 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:watchdog)
(thread-sleep! 0.05) ;; delay for startup
(let ((legacy-sync (common:run-sync?))
(debug-mode (debug:debug-mode 1))
(last-time (current-seconds))
(this-wd-num (begin (mutex-lock! *wdnum*mutex) (let ((x *wdnum*)) (set! *wdnum* (add1 *wdnum*)) (mutex-unlock! *wdnum*mutex) x)))
)
(debug:print-info 0 *default-log-port* "watchdog starting. legacy-sync is " legacy-sync" pid="(current-process-id)" this-wd-num="this-wd-num)
(if (and legacy-sync (not *time-to-exit*))
(let ((dbstruct (db:setup)))
(debug:print-info 0 *default-log-port* "Server running, periodic sync started.")
(let loop ()
(BB> "watchdog loop. pid="(current-process-id)" this-wd-num="this-wd-num" *time-to-exit*="*time-to-exit*)
;; sync for filesystem local db writes
;;
(mutex-lock! *db-multi-sync-mutex*)
(let* ((need-sync (>= *db-last-access* *db-last-sync*)) ;; no sync since last write
(sync-in-progress *db-sync-in-progress*)
(should-sync (and (not *time-to-exit*)
(> (- (current-seconds) *db-last-sync*) 5))) ;; sync every five seconds minimum
(will-sync (and (or need-sync should-sync)
(not sync-in-progress)))
(start-time (current-seconds)))
;; (debug:print-info 0 *default-log-port* "need-sync: " need-sync " sync-in-progress: " sync-in-progress " should-sync: " should-sync " will-sync: " will-sync)
(if will-sync (set! *db-sync-in-progress* #t))
(mutex-unlock! *db-multi-sync-mutex*)
(if will-sync
|
| ︙ | ︙ | |||
607 608 609 610 611 612 613 614 615 616 617 618 | (set! last-time start-time) (debug:print-info 4 *default-log-port* "timestamp -> " (seconds->time-string (current-seconds)) ", time since start -> " (seconds->hr-min-sec (- (current-seconds) *time-zero*)))))) ;; keep going unless time to exit ;; (if (not *time-to-exit*) (let delay-loop ((count 0)) (if (and (not *time-to-exit*) (< count 4)) ;; was 11, changing to 4. (begin (thread-sleep! 1) (delay-loop (+ count 1)))) | > > | | | 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 |
(set! last-time start-time)
(debug:print-info 4 *default-log-port* "timestamp -> " (seconds->time-string (current-seconds)) ", time since start -> " (seconds->hr-min-sec (- (current-seconds) *time-zero*))))))
;; keep going unless time to exit
;;
(if (not *time-to-exit*)
(let delay-loop ((count 0))
(BB> "delay-loop top; count="count" pid="(current-process-id)" this-wd-num="this-wd-num" *time-to-exit*="*time-to-exit*)
(if (and (not *time-to-exit*)
(< count 4)) ;; was 11, changing to 4.
(begin
(thread-sleep! 1)
(delay-loop (+ count 1))))
(if (not *time-to-exit*) (loop))))
(if (common:low-noise-print 30)
(debug:print-info 0 *default-log-port* "Exiting watchdog timer, *time-to-exit* = " *time-to-exit*" pid="(current-process-id)" this-wd-num="this-wd-num)))))))
(define (std-exit-procedure)
(on-exit (lambda () 0))
(let ((no-hurry (if *time-to-exit* ;; hurry up
#f
(begin
(set! *time-to-exit* #t)
|
| ︙ | ︙ | |||
639 640 641 642 643 644 645 |
(sqlite3:interrupt! db)
(sqlite3:finalize! db #t)
;; (vector-set! *task-db* 0 #f)
(set! *task-db* #f)))))
(if (and *runremote*
(remote-conndat *runremote*))
(begin
| | | 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 |
(sqlite3:interrupt! db)
(sqlite3:finalize! db #t)
;; (vector-set! *task-db* 0 #f)
(set! *task-db* #f)))))
(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
|
| ︙ | ︙ |
Modified http-transport.scm from [ec36961174] to [285317a3d3].
| ︙ | ︙ | |||
484 485 486 487 488 489 490 | ;; (if (and (<= rem-time 4) ;; (> rem-time 0)) ;; (thread-sleep! rem-time) ;; (thread-sleep! 4))) ;; fallback for if the math is changed ... (define (http-transport:server-shutdown server-id port) (let ((tdbdat (tasks:open-db))) | | | 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 |
;; (if (and (<= rem-time 4)
;; (> rem-time 0))
;; (thread-sleep! rem-time)
;; (thread-sleep! 4))) ;; fallback for if the math is changed ...
(define (http-transport:server-shutdown server-id port)
(let ((tdbdat (tasks:open-db)))
(debug:print-info 0 *default-log-port* "Starting to shutdown the server. pid="(current-process-id))
;;
;; start_shutdown
;;
(tasks:server-set-state! (db:delay-if-busy tdbdat) server-id "shutting-down")
(set! *time-to-exit* #t) ;; tell on-exit to be fast as we've already cleaned up
(portlogger:open-run-close portlogger:set-port port "released")
(thread-sleep! 5)
|
| ︙ | ︙ |
Modified megatest.scm from [ea009aa7f2] to [e7342b90ae].
| ︙ | ︙ | |||
348 349 350 351 352 353 354 | (if targ (setenv "MT_TARGET" targ))) ;; The watchdog is to keep an eye on things like db sync etc. ;; (define *watchdog* (make-thread common:watchdog "Watchdog thread")) (thread-start! *watchdog*) | | | 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 |
(if targ (setenv "MT_TARGET" targ)))
;; The watchdog is to keep an eye on things like db sync etc.
;;
(define *watchdog* (make-thread common:watchdog "Watchdog thread"))
(thread-start! *watchdog*)
;; (BB> "thread-start! watchdog")
(if (args:get-arg "-log")
(let ((oup (open-output-file (args:get-arg "-log"))))
(debug:print-info 0 *default-log-port* "Sending log output to " (args:get-arg "-log"))
(set! *default-log-port* oup)))
(if (or (args:get-arg "-h")
(args:get-arg "-help")
|
| ︙ | ︙ | |||
1988 1989 1990 1991 1992 1993 1994 |
;;======================================================================
;; Exit and clean up
;;======================================================================
(if (not *didsomething*)
(debug:print 0 *default-log-port* help))
| | | 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 |
;;======================================================================
;; Exit and clean up
;;======================================================================
(if (not *didsomething*)
(debug:print 0 *default-log-port* help))
(BB> "thread-join! watchdog")
(thread-join! *watchdog*)
(set! *time-to-exit* #t)
(if (not (eq? *globalexitstatus* 0))
(if (or (args:get-arg "-run")(args:get-arg "-runtests")(args:get-arg "-runall"))
(begin
(debug:print 0 *default-log-port* "NOTE: Subprocesses with non-zero exit code detected: " *globalexitstatus*)
(exit 0))
(case *globalexitstatus*
((0)(exit 0))
((1)(exit 1))
((2)(exit 2))
(else (exit 3)))))
|
Modified rmt.scm from [7e5abf90aa] to [c70f311cf0].
| ︙ | ︙ | |||
84 85 86 87 88 89 90 91 92 93 94 95 96 97 |
(rmt:send-receive cmd rid params attemptnum: attemptnum))
;; on homehost and this is a read
((and (cdr (remote-hh-dat *runremote*)) ;; on homehost
(member cmd api:read-only-queries)) ;; this is a read
(mutex-unlock! *rmt-mutex*)
(debug:print-info 12 *default-log-port* "rmt:send-receive, case 3")
(rmt:open-qry-close-locally cmd 0 params))
;; on homehost and this is a write, we already have a server
((and (cdr (remote-hh-dat *runremote*)) ;; on homehost
(not (member cmd api:read-only-queries)) ;; this is a write
(remote-server-url *runremote*)) ;; have a server
(mutex-unlock! *rmt-mutex*)
(debug:print-info 12 *default-log-port* "rmt:send-receive, case 4")
(rmt:open-qry-close-locally cmd 0 params))
| > > > > > > > > > > > > > | | | | | | > > | 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 |
(rmt:send-receive cmd rid params attemptnum: attemptnum))
;; on homehost and this is a read
((and (cdr (remote-hh-dat *runremote*)) ;; on homehost
(member cmd api:read-only-queries)) ;; this is a read
(mutex-unlock! *rmt-mutex*)
(debug:print-info 12 *default-log-port* "rmt:send-receive, case 3")
(rmt:open-qry-close-locally cmd 0 params))
;; on homehost and this is a write, we already have a server, but server has died
((and (cdr (remote-hh-dat *runremote*)) ;; on homehost
(not (member cmd api:read-only-queries)) ;; this is a write
(remote-server-url *runremote*) ;; have a server
(not (server:read-dotserver *toppath*))) ;; server has died.
(set! *runremote* #f)
(mutex-unlock! *rmt-mutex*)
(debug:print-info 12 *default-log-port* "rmt:send-receive, case 4.1")
(rmt:send-receive cmd rid params attemptnum: attemptnum))
;; on homehost and this is a write, we already have a server
((and (cdr (remote-hh-dat *runremote*)) ;; on homehost
(not (member cmd api:read-only-queries)) ;; this is a write
(remote-server-url *runremote*)) ;; have a server
(mutex-unlock! *rmt-mutex*)
(debug:print-info 12 *default-log-port* "rmt:send-receive, case 4")
(rmt:open-qry-close-locally cmd 0 params))
;; commented by bb; this was blocking server passive start on write on homehost (case 5)
;; ;; on homehost and this is a write, we have a server (we know because case 4 checked)
;; ((and (cdr (remote-hh-dat *runremote*)) ;; on homehost
;; (not (member cmd api:read-only-queries)))
;; (mutex-unlock! *rmt-mutex*)
;; (debug:print-info 12 *default-log-port* "rmt:send-receive, case 4.1")
;; (rmt:open-qry-close-locally cmd 0 params))
;; no server contact made and this is a write, passively start a server
((and (not (remote-server-url *runremote*))
(not (member cmd api:read-only-queries)))
(debug:print-info 12 *default-log-port* "rmt:send-receive, case 5")
(let ((serverconn (server:read-dotserver *toppath*))) ;; (server:check-if-running *toppath*))) ;; Do NOT want to run server:check-if-running - very expensive to do for every write call
(if serverconn
(remote-server-url-set! *runremote* serverconn) ;; the string can be consumed by the client setup if needed
|
| ︙ | ︙ | |||
347 348 349 350 351 352 353 354 355 356 357 358 359 360 |
(rmt:send-receive 'get-key-val-pairs run-id (list run-id)))
(define (rmt:get-keys)
(if *db-keys* *db-keys*
(let ((res (rmt:send-receive 'get-keys #f '())))
(set! *db-keys* res)
res)))
;; we don't reuse run-id's (except possibly *after* a db cleanup) so it is safe
;; to cache the resuls in a hash
;;
(define (rmt:get-key-vals run-id)
(or (hash-table-ref/default *keyvals* run-id #f)
(let ((res (rmt:send-receive 'get-key-vals #f (list run-id))))
| > > > > > | 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 |
(rmt:send-receive 'get-key-val-pairs run-id (list run-id)))
(define (rmt:get-keys)
(if *db-keys* *db-keys*
(let ((res (rmt:send-receive 'get-keys #f '())))
(set! *db-keys* res)
res)))
(define (rmt:get-keys-write) ;; dummy query to force server start
(let ((res (rmt:send-receive 'get-keys-write #f '())))
(set! *db-keys* res)
res))
;; we don't reuse run-id's (except possibly *after* a db cleanup) so it is safe
;; to cache the resuls in a hash
;;
(define (rmt:get-key-vals run-id)
(or (hash-table-ref/default *keyvals* run-id #f)
(let ((res (rmt:send-receive 'get-key-vals #f (list run-id))))
|
| ︙ | ︙ |