Changes In Branch v1.6569-refactor-server-key-chk Through [70f5d2ff31] Excluding Merge-Ins
This is equivalent to a diff from 86a3d1148e to 70f5d2ff31
2021-01-29
| ||
10:28 | changed info debug msgs with keyword failed check-in: 577abe1ca7 user: pjhatwal tags: v1.6569-refactor-server-key-chk | |
2021-01-28
| ||
18:40 | fixes to crashes with server id mismatch check-in: 70f5d2ff31 user: pjhatwal tags: v1.6569-refactor-server-key-chk | |
2021-01-27
| ||
15:30 | fix for crash with error (match) no matching pattern check-in: 4c4d63c3e0 user: pjhatwal tags: v1.6569-refactor-server-key-chk | |
2021-01-14
| ||
14:55 | check server-key on every request server gets check-in: f74b755ed8 user: pjhatwal tags: v1.6569-refactor-server-key-chk | |
2021-01-08
| ||
12:12 | cherrypicked changes from 1.65-archive Leaf check-in: b72620bfbb user: pjhatwal tags: v1.6569-pjhatwal-mess | |
11:46 | Create new branch named "v1.6569-refactor-archive-structure-change" check-in: bef664ccc1 user: pjhatwal tags: v1.6569-refactor-archive-structure-change | |
11:42 | enable custom value for max delay between archive time and test last update time Leaf check-in: 86a3d1148e user: pjhatwal tags: v1.6569-refactor | |
2020-09-18
| ||
17:30 | added check for file existence before file delete ==/14/1.9/WARN/orion,mars/== NOTE: This is the last v1.65 before the split off. I.e code from before this point IS in the far future v1.65 branch. Code from this point to that branch might NOT be in the branch. check-in: 2769e4b7c9 user: mmgraham tags: v1.65, v1.6569 | |
Modified api.scm from [4fa67bb6bd] to [68ac71805c].
1 2 3 4 5 6 7 | ;;====================================================================== ;; Copyright 2006-2013, Matthew Welland. ;; ;; This file is part of Megatest. ;; ;; Megatest is free software: you can redistribute it and/or modify ;; it under the terms of the GNU General Public License as published by | > > | 1 2 3 4 5 6 7 8 9 | ;;====================================================================== ;; Copyright 2006-2013, Matthew Welland. ;; ;; This file is part of Megatest. ;; ;; Megatest is free software: you can redistribute it and/or modify ;; it under the terms of the GNU General Public License as published by |
︙ | ︙ | |||
374 375 376 377 378 379 380 | ;; http-server send-response ;; api:process-request ;; db:* ;; ;; NB// Runs on the server as part of the server loop ;; (define (api:process-request dbstruct $) ;; the $ is the request vars proc | | > | > > > > | | | > | | | | | | | | | | | | | | > > > | 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 | ;; http-server send-response ;; api:process-request ;; db:* ;; ;; NB// Runs on the server as part of the server loop ;; (define (api:process-request dbstruct $) ;; the $ is the request vars proc (debug:print 4 *default-log-port* "server-id:" *server-id*) (let* ((cmd ($ 'cmd)) (paramsj ($ 'params)) (key ($ 'key)) (params (db:string->obj paramsj transport: 'http))) ;; incoming data from the POST (or is it a GET?) (debug:print 0 *default-log-port* "cmd:" cmd " with params " params "key " key) (if (equal? key *server-id*) (begin (set! *api-process-request-count* (+ *api-process-request-count* 1)) (let* ((resdat (api:execute-requests dbstruct (vector cmd params))) ;; process the request, resdat = #( flag result ) (success (vector-ref resdat 0)) (res (vector-ref resdat 1))) ;; (vector flag payload), get the payload, ignore the flag (why?) (debug:print 0 *default-log-port* "res:" res) (if (not success) (debug:print 0 *default-log-port* "ERROR: success flag is #f for " cmd " with params " params)) (if (> *api-process-request-count* *max-api-process-requests*) (set! *max-api-process-requests* *api-process-request-count*)) (set! *api-process-request-count* (- *api-process-request-count* 1)) ;; This can be here but needs controls to ensure it doesn't run more than every 4 seconds ;; (rmt:dat->json-str ;; (if (or (string? res) ;; (list? res) ;; (number? res) ;; (boolean? res)) ;; res ;; (list "ERROR, not string, list, number or boolean" 1 cmd params res))))) (db:obj->string res transport: 'http))) (begin (debug:print 0 *default-log-port* "Server refused to process request. Sever id mismatch. recived " key " expected: " *server-id* ".\nOther arguments recived: cmd=" cmd " params = " params) (db:obj->string (conc "Server refused to process request server-id mismatch: " key ", " *server-id*) transport: 'http))))) |
Modified client.scm from [9da8d7475d] to [dc4c7b41e8].
︙ | ︙ | |||
86 87 88 89 90 91 92 | ;; through them searching for a good one. ;; (let* ((server-dat (server:get-rand-best areapath)) ;; (server:get-first-best areapath)) (runremote (or area-dat *runremote*))) (if (not server-dat) ;; no server found (client:setup-http areapath remaining-tries: (- remaining-tries 1)) (let ((host (cadr server-dat)) | | > > | > > > > > | | | 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 | ;; through them searching for a good one. ;; (let* ((server-dat (server:get-rand-best areapath)) ;; (server:get-first-best areapath)) (runremote (or area-dat *runremote*))) (if (not server-dat) ;; no server found (client:setup-http areapath remaining-tries: (- remaining-tries 1)) (let ((host (cadr server-dat)) (port (caddr server-dat)) (server-id (caddr (cddr server-dat)))) (debug:print-info 4 *default-log-port* "client:setup server-dat=" server-dat ", remaining-tries=" remaining-tries) (if (and (not area-dat) (not *runremote*)) (begin (set! *runremote* (make-remote)) (let* ((server-info (remote-server-info *runremote*))) (if server-info (begin (remote-server-url-set! *runremote* (server:record->url server-info)) (remote-server-id-set! *runremote* (server:record->id server-info))))))) (if (and host port server-id) (let* ((start-res (case *transport-type* ((http)(http-transport:client-connect host port server-id)))) (ping-res (case *transport-type* ((http)(rmt:login-no-auto-client-setup start-res))))) (if (and start-res ping-res) (let ((runremote (or area-dat *runremote*))) ;; it might have been generated only a few statements ago (remote-conndat-set! runremote start-res) ;; (hash-table-set! runremote run-id start-res) (debug:print-info 2 *default-log-port* "connected to " (http-transport:server-dat-make-url start-res)) |
︙ | ︙ |
Modified common.scm from [33c7316880] to [bf0a0a25ad].
︙ | ︙ | |||
294 295 296 297 298 299 300 | (else "FAIL"))) (define (common:logpro-exit-code->test-status exit-code) (status-sym->string (common:logpro-exit-code->status-sym exit-code))) (defstruct remote (hh-dat (common:get-homehost)) ;; homehost record ( addr . hhflag ) | > > | | 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 | (else "FAIL"))) (define (common:logpro-exit-code->test-status exit-code) (status-sym->string (common:logpro-exit-code->status-sym exit-code))) (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 |
︙ | ︙ |
Modified db.scm from [fb3a18f52f] to [f2d817bbad].
︙ | ︙ | |||
3213 3214 3215 3216 3217 3218 3219 | (if newstatus (sqlite3:execute db "UPDATE tests SET status=? WHERE id=?;" newstatus test-id)) (if newcomment (sqlite3:execute db "UPDATE tests SET comment=? WHERE id=?;" newcomment ;; (sdb:qry 'getid newcomment) test-id)))))) (mt:process-triggers dbstruct run-id test-id newstate newstatus)) ;; NEW BEHAVIOR: Count tests running in all runs! ;; | | | | | | 3213 3214 3215 3216 3217 3218 3219 3220 3221 3222 3223 3224 3225 3226 3227 3228 3229 3230 | (if newstatus (sqlite3:execute db "UPDATE tests SET status=? WHERE id=?;" newstatus test-id)) (if newcomment (sqlite3:execute db "UPDATE tests SET comment=? WHERE id=?;" newcomment ;; (sdb:qry 'getid newcomment) test-id)))))) (mt:process-triggers dbstruct run-id test-id newstate newstatus)) ;; NEW BEHAVIOR: Count tests running in all runs! ;; (define (db:get-count-tests-running dbstruct run-id) ;; fastmode) (let* ((qry ;; (if fastmode ;; "SELECT count(id) FROM tests WHERE state in ('RUNNING','LAUNCHED','REMOTEHOSTSTART') AND NOT (uname = 'n/a' AND item_path = '') LIMIT 1;" "SELECT count(id) FROM tests WHERE state in ('RUNNING','LAUNCHED','REMOTEHOSTSTART') AND NOT (uname = 'n/a' AND item_path = '');")) ;; ) (db:with-db dbstruct run-id #f (lambda (db) (let* ((stmth (db:get-cache-stmth dbstruct db qry))) (sqlite3:first-result stmth)))))) |
︙ | ︙ | |||
3243 3244 3245 3246 3247 3248 3249 | ;; "SELECT count(id) FROM tests WHERE state in ('RUNNING','LAUNCHED','REMOTEHOSTSTART') AND run_id NOT IN (SELECT id FROM runs WHERE state='deleted') AND NOT (uname = 'n/a' AND item_path = '');") "SELECT count(id) FROM tests WHERE state in ('RUNNING','REMOTEHOSTSTART','LAUNCHED') AND run_id=?;" run-id)))) ;; NOT IN (SELECT id FROM runs WHERE state='deleted');") ;; NEW BEHAVIOR: Look only at single run with run-id ;; ;; (define (db:get-running-stats dbstruct run-id) | | | | | | 3243 3244 3245 3246 3247 3248 3249 3250 3251 3252 3253 3254 3255 3256 3257 3258 3259 3260 | ;; "SELECT count(id) FROM tests WHERE state in ('RUNNING','LAUNCHED','REMOTEHOSTSTART') AND run_id NOT IN (SELECT id FROM runs WHERE state='deleted') AND NOT (uname = 'n/a' AND item_path = '');") "SELECT count(id) FROM tests WHERE state in ('RUNNING','REMOTEHOSTSTART','LAUNCHED') AND run_id=?;" run-id)))) ;; NOT IN (SELECT id FROM runs WHERE state='deleted');") ;; NEW BEHAVIOR: Look only at single run with run-id ;; ;; (define (db:get-running-stats dbstruct run-id) (define (db:get-count-tests-running-for-run-id dbstruct run-id) ;; fastmode) (let* ((qry ;; (if fastmode ;; "SELECT count(id) FROM tests WHERE state in ('RUNNING','LAUNCHED','REMOTEHOSTSTART') AND run_id=? LIMIT 1;" "SELECT count(id) FROM tests WHERE state in ('RUNNING','LAUNCHED','REMOTEHOSTSTART') AND run_id=?;")) ;; ) (db:with-db dbstruct run-id #f (lambda (db) (let* ((stmth (db:get-cache-stmth dbstruct db qry))) (sqlite3:first-result stmth run-id)))))) |
︙ | ︙ | |||
4340 4341 4342 4343 4344 4345 4346 | flush sync set-verbosity killserver )) (define (db:login dbstruct calling-path calling-version client-signature) | | > | | 4340 4341 4342 4343 4344 4345 4346 4347 4348 4349 4350 4351 4352 4353 4354 4355 4356 4357 4358 4359 4360 4361 4362 | flush sync set-verbosity killserver )) (define (db:login dbstruct calling-path calling-version client-signature) (cond ((not (equal? calling-path *toppath*)) (list #f "Login failed due to mismatch paths: " calling-path ", " *toppath*)) ;; ((not (equal? *run-id* run-id)) ;; (list #f "Login failed due to mismatch run-id: " run-id ", " *run-id*)) ((not (equal? megatest-version calling-version)) (list #t (conc "Login warning due to mismatch megatest version: " calling-version ", " megatest-version))) (else (hash-table-set! *logged-in-clients* client-signature (current-seconds)) '(#t "successful login")))) (define (db:general-call dbstruct stmtname params) (let ((query (let ((q (alist-ref (if (string? stmtname) (string->symbol stmtname) stmtname) |
︙ | ︙ |
Modified dcommon.scm from [0db7864f6b] to [dbcf309f44].
︙ | ︙ | |||
711 712 713 714 715 716 717 | ;; (iup:attribute-set! servers-matrix (conc "0:" colnum) colname) ;; (set! colnum (+ 1 colnum))) ;; colnames) (set! rownum 1) (for-each (lambda (server) (set! colnum 0) | | | 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 | ;; (iup:attribute-set! servers-matrix (conc "0:" colnum) colname) ;; (set! colnum (+ 1 colnum))) ;; colnames) (set! rownum 1) (for-each (lambda (server) (set! colnum 0) (match-let (((mod-time host port start-time server-id pid) server)) (let* ((uptime (- (current-seconds) mod-time)) (runtime (if start-time (- mod-time start-time) 0)) (vals (list "-" ;; (vector-ref server 0) ;; Id "-" ;; (vector-ref server 9) ;; MT-Ver |
︙ | ︙ |
Modified http-transport.scm from [67489ed9ab] to [2202b22e9f].
︙ | ︙ | |||
244 245 246 247 248 249 250 | (http-transport:server-dat-get-api-req serverdat) (begin (debug:print 0 *default-log-port* "FATAL ERROR: http-transport:client-api-send-receive called with no server info") (exit 1)))) (res (vector #f "uninitialized")) (success #t) (sparams (db:obj->string params transport: 'http)) | | > > > > > | > | 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 | (http-transport:server-dat-get-api-req serverdat) (begin (debug:print 0 *default-log-port* "FATAL ERROR: http-transport:client-api-send-receive called with no server info") (exit 1)))) (res (vector #f "uninitialized")) (success #t) (sparams (db:obj->string params transport: 'http)) (runremote (or area-dat *runremote*)) (server-id (if (vector? serverdat) (http-transport:server-dat-get-server-id serverdat) (begin (debug:print 0 *default-log-port* "FATAL ERROR: http-transport:client-api-send-receive called with no server info") (exit 1))))) (debug:print-info 11 *default-log-port* "cmd=" cmd " fullurl=" fullurl " server-id=" server-id " current time:" (current-seconds)) ;; set up the http-client here (max-retry-attempts 1) ;; consider all requests indempotent (retry-request? (lambda (request) #f)) ;; send the data and get the response ;; extract the needed info from the http data and |
︙ | ︙ | |||
271 272 273 274 275 276 277 | (msg ((condition-property-accessor 'exn 'message) exn))) (set! success #f) (if (debug:debug-mode 1) (debug:print-info 0 *default-log-port* "couldn't talk to server, trying again ...") (begin (debug:print 0 *default-log-port* "WARNING: failure in with-input-from-request to " fullurl ".") (debug:print 0 *default-log-port* " message: " msg ", exn=" exn) | | | > < | > > > > > > | | 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 | (msg ((condition-property-accessor 'exn 'message) exn))) (set! success #f) (if (debug:debug-mode 1) (debug:print-info 0 *default-log-port* "couldn't talk to server, trying again ...") (begin (debug:print 0 *default-log-port* "WARNING: failure in with-input-from-request to " fullurl ".") (debug:print 0 *default-log-port* " message: " msg ", exn=" exn) (debug:print 0 *default-log-port* " cmd: " cmd " params: " params " key:" (or server-id "thekey")) (debug:print 0 *default-log-port* " call-chain: " call-chain))) (if runremote (remote-conndat-set! runremote #f)) ;; Killing associated server to allow clean retry.") ;; (tasks:kill-server-run-id run-id) ;; better to kill the server in the logic that called this routine? (mutex-unlock! *http-mutex*) ;;; (signal (make-composite-condition ;;; (make-property-condition 'commfail 'message "failed to connect to server"))) ;;; "communications failed" (db:obj->string #f)) (with-input-from-request ;; was dat fullurl (list (cons 'key (or server-id "thekey")) (cons 'cmd cmd) (cons 'params sparams)) read-string)) transport: 'http) 0)) ;; added this speculatively ;; Shouldn't this be a call to the managed call-all-connections stuff above? (close-all-connections!) (mutex-unlock! *http-mutex*) )) (time-out (lambda () (thread-sleep! 45) (debug:print 0 *default-log-port* "WARNING: send-receive took more than 45 seconds!!") #f)) (th1 (make-thread send-recieve "with-input-from-request")) (th2 (make-thread time-out "time out"))) (thread-start! th1) (thread-start! th2) (thread-join! th1) (vector-set! res 0 success) (thread-terminate! th2) (if (vector? res) (if (vector-ref res 0) ;; this is the first flag or the second flag? (let* ((res-dat (vector-ref res 1))) (if (and (string? res-dat) (string-contains res-dat "server-id mismatch")) (signal (make-composite-condition (make-property-condition 'servermismatch 'message (vector-ref res 1)))) res)) ;; this is the *inner* vector? seriously? why? (if (debug:debug-mode 11) (let ((call-chain (get-call-chain))) ;; note: this code also called in nmsg-transport - consider consolidating it (print-call-chain (current-error-port)) (debug:print-error 11 *default-log-port* "error above occured at server, res=" res) ;; " message: " ((condition-property-accessor 'exn 'message) exn)) (debug:print 11 *default-log-port* " server call chain:") (pp (vector-ref res 1) (current-error-port)) (signal (vector-ref res 0))) |
︙ | ︙ | |||
348 349 350 351 352 353 354 | (define (make-http-transport:server-dat)(make-vector 6)) (define (http-transport:server-dat-get-iface vec) (vector-ref vec 0)) (define (http-transport:server-dat-get-port vec) (vector-ref vec 1)) (define (http-transport:server-dat-get-api-uri vec) (vector-ref vec 2)) (define (http-transport:server-dat-get-api-url vec) (vector-ref vec 3)) (define (http-transport:server-dat-get-api-req vec) (vector-ref vec 4)) (define (http-transport:server-dat-get-last-access vec) (vector-ref vec 5)) | | > | | | 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 | (define (make-http-transport:server-dat)(make-vector 6)) (define (http-transport:server-dat-get-iface vec) (vector-ref vec 0)) (define (http-transport:server-dat-get-port vec) (vector-ref vec 1)) (define (http-transport:server-dat-get-api-uri vec) (vector-ref vec 2)) (define (http-transport:server-dat-get-api-url vec) (vector-ref vec 3)) (define (http-transport:server-dat-get-api-req vec) (vector-ref vec 4)) (define (http-transport:server-dat-get-last-access vec) (vector-ref vec 5)) ;(define (http-transport:server-dat-get-socket vec) (vector-ref vec 6)) (define (http-transport:server-dat-get-server-id vec) (vector-ref vec 6)) (define (http-transport:server-dat-make-url vec) (if (and (http-transport:server-dat-get-iface vec) (http-transport:server-dat-get-port vec)) (conc "http://" (http-transport:server-dat-get-iface vec) ":" (http-transport:server-dat-get-port vec)) #f)) (define (http-transport:server-dat-update-last-access vec) (if (vector? vec) (vector-set! vec 5 (current-seconds)) (begin (print-call-chain (current-error-port)) (debug:print-error 0 *default-log-port* "call to http-transport:server-dat-update-last-access with non-vector!!")))) ;; ;; connect ;; (define (http-transport:client-connect iface port server-id) (let* ((api-url (conc "http://" iface ":" port "/api")) (api-uri (uri-reference (conc "http://" iface ":" port "/api"))) (api-req (make-request method: 'POST uri: api-uri)) (server-dat (vector iface port api-uri api-url api-req (current-seconds) server-id))) server-dat)) ;; run http-transport:keep-running in a parallel thread to monitor that the db is being ;; used and to shutdown after sometime if it is not. ;; (define (http-transport:keep-running) ;; if none running or if > 20 seconds since |
︙ | ︙ | |||
473 474 475 476 477 478 479 | (if (not (equal? sdat (list iface port))) (let ((new-iface (car sdat)) (new-port (cadr sdat))) (debug:print-info 0 *default-log-port* "WARNING: interface changed, refreshing iface and port info") (set! iface new-iface) (set! port new-port) | > > > | > > > | | 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 | (if (not (equal? sdat (list iface port))) (let ((new-iface (car sdat)) (new-port (cadr sdat))) (debug:print-info 0 *default-log-port* "WARNING: interface changed, refreshing iface and port info") (set! iface new-iface) (set! port new-port) (if (not *server-id*) (set! *server-id* (server:mk-signature))) (debug:print 0 *default-log-port* (current-seconds) (current-directory) (current-process-id) (argv)) (debug:print 0 *default-log-port* "SERVER STARTED: " iface ":" port " AT " (current-seconds) " server-id: " *server-id*) (flush-output *default-log-port*))) ;; Transfer *db-last-access* to last-access to use in checking that we are still alive (mutex-lock! *heartbeat-mutex*) (set! last-access *db-last-access*) (mutex-unlock! *heartbeat-mutex*) (if (common:low-noise-print 120 (conc "server running on " iface ":" port)) (begin (if (not *server-id*) (set! *server-id* (server:mk-signature))) (debug:print 0 *default-log-port* (current-seconds) (current-directory) (current-process-id) (argv)) (debug:print 0 *default-log-port* "SERVER STARTED: " iface ":" port " AT " (current-seconds) " server-id: " *server-id*) (flush-output *default-log-port*))) (if (common:low-noise-print 60 "dbstats") (begin (debug:print 0 *default-log-port* "Server stats:") (db:print-current-query-stats))) (let* ((hrs-since-start (/ (- (current-seconds) server-start-time) 3600))) (cond |
︙ | ︙ |
Modified launch.scm from [d0067277fa] to [e8093b3e63].
︙ | ︙ | |||
729 730 731 732 733 734 735 | ;; 100% COMPLETED/ (PASS,FAIL,ABORT etc.) ==> COMPLETED / X where X is same as itemized rollup ;; > 3 RUNNING with not test_dead do nothing (run should already be RUNNING/ na ;; > 0 RUNNING and test_dead then send KILLREQ ==> COMPLETED ;; 0 RUNNING ==> this is actually the first condition, should not get here (define (launch:end-of-run-check run-id ) (let* ((not-completed-cnt (rmt:get-not-completed-cnt run-id)) | | | 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 | ;; 100% COMPLETED/ (PASS,FAIL,ABORT etc.) ==> COMPLETED / X where X is same as itemized rollup ;; > 3 RUNNING with not test_dead do nothing (run should already be RUNNING/ na ;; > 0 RUNNING and test_dead then send KILLREQ ==> COMPLETED ;; 0 RUNNING ==> this is actually the first condition, should not get here (define (launch:end-of-run-check run-id ) (let* ((not-completed-cnt (rmt:get-not-completed-cnt run-id)) (running-cnt (rmt:get-count-tests-running-for-run-id run-id)) (all-test-launched (rmt:get-var (conc "lunch-complete-" run-id))) (current-state (rmt:get-run-state run-id)) (current-status (rmt:get-run-status run-id))) ;;get-vars run-id to query metadata table to check if all completed. if all-test-launched = yes then only not-completed-cnt = 0 means everyting is completed if no entry found in the table do nothing (debug:print 0 *default-log-port* "Running test cnt :" running-cnt) (rmt:set-state-status-and-roll-up-run run-id current-state current-status) (runs:update-junit-test-reporter-xml run-id) |
︙ | ︙ |
Modified megatest-version.scm from [f253e5978b] to [69c8b1f2d8].
︙ | ︙ | |||
16 17 18 19 20 21 22 | ;; along with Megatest. If not, see <http://www.gnu.org/licenses/>. ;; Always use two or four digit decimal ;; 1.01, 1.02...1.10,1.11,1.1101 ... 1.99,2.00.. ;; (declare (unit megatest-version)) | | | 16 17 18 19 20 21 22 23 | ;; along with Megatest. If not, see <http://www.gnu.org/licenses/>. ;; Always use two or four digit decimal ;; 1.01, 1.02...1.10,1.11,1.1101 ... 1.99,2.00.. ;; (declare (unit megatest-version)) (define megatest-version 1.6581) |
Modified megatest.scm from [0e58f17e0f] to [e69eff1234].
︙ | ︙ | |||
850 851 852 853 854 855 856 | (if out-file (close-output-port out-port)) (exit) ;; yes, bending the rules here - need to exit since this is a utility )) (if (args:get-arg "-ping") (let* ((server-id (string->number (args:get-arg "-ping"))) ;; extract run-id (i.e. no ":" (host:port (args:get-arg "-ping"))) | | | 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 | (if out-file (close-output-port out-port)) (exit) ;; yes, bending the rules here - need to exit since this is a utility )) (if (args:get-arg "-ping") (let* ((server-id (string->number (args:get-arg "-ping"))) ;; extract run-id (i.e. no ":" (host:port (args:get-arg "-ping"))) (server:ping (or server-id host:port) #f do-exit: #t))) ;;====================================================================== ;; Capture, save and manipulate environments ;;====================================================================== ;; NOTE: Keep these above the section where the server or client code is setup |
︙ | ︙ |
Modified rmt.scm from [39d97c528a] to [bcbb74efcc].
︙ | ︙ | |||
98 99 100 101 102 103 104 105 106 107 108 109 110 111 | ;; DOT INIT_RUNREMOTE; // leaving off - doesn't really add to the clarity ;; DOT MUTEXLOCK -> INIT_RUNREMOTE [label="no remote?"]; ;; DOT INIT_RUNREMOTE -> MUTEXLOCK; ;; ensure we have a record for our connection for given area (if (not runremote) ;; can remove this one. should never get here. (begin (set! *runremote* (make-remote)) (set! runremote *runremote*))) ;; new runremote will come from this on next iteration ;; DOT SET_HOMEHOST; // leaving off - doesn't really add to the clarity ;; DOT MUTEXLOCK -> SET_HOMEHOST [label="no homehost?"]; ;; DOT SET_HOMEHOST -> MUTEXLOCK; ;; ensure we have a homehost record (if (not (pair? (remote-hh-dat runremote))) ;; not on homehost | > > > > > | 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 | ;; DOT INIT_RUNREMOTE; // leaving off - doesn't really add to the clarity ;; DOT MUTEXLOCK -> INIT_RUNREMOTE [label="no remote?"]; ;; DOT INIT_RUNREMOTE -> MUTEXLOCK; ;; ensure we have a record for our connection for given area (if (not runremote) ;; can remove this one. should never get here. (begin (set! *runremote* (make-remote)) (let* ((server-info (remote-server-info *runremote*))) (if server-info (begin (remote-server-url-set! *runremote* (server:record->url server-info)) (remote-server-id-set! *runremote* (server:record->id server-info))))) (set! runremote *runremote*))) ;; new runremote will come from this on next iteration ;; DOT SET_HOMEHOST; // leaving off - doesn't really add to the clarity ;; DOT MUTEXLOCK -> SET_HOMEHOST [label="no homehost?"]; ;; DOT SET_HOMEHOST -> MUTEXLOCK; ;; ensure we have a homehost record (if (not (pair? (remote-hh-dat runremote))) ;; not on homehost |
︙ | ︙ | |||
172 173 174 175 176 177 178 | ;;DOT CASE6 [label="init\nremote"]; ;;DOT MUTEXLOCK -> CASE6 [label="on homehost,\nwrite query,\nhave server,\ncan't reach it"]; {rank=same "case 6" CASE6}; ;;DOT CASE6 -> "rmt:send-receive"; ;; 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 | | > > > > > | 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 | ;;DOT CASE6 [label="init\nremote"]; ;;DOT MUTEXLOCK -> CASE6 [label="on homehost,\nwrite query,\nhave server,\ncan't reach it"]; {rank=same "case 6" CASE6}; ;;DOT CASE6 -> "rmt:send-receive"; ;; 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:ping (remote-server-url runremote) (remote-server-id runremote)))) ;; server has died. NOTE: this is not a cheap call! Need better approach. (set! *runremote* (make-remote)) (let* ((server-info (remote-server-info *runremote*))) (if server-info (begin (remote-server-url-set! *runremote* (server:record->url server-info)) (remote-server-id-set! *runremote* (server:record->id server-info))))) (remote-force-server-set! runremote (common:force-server?)) (mutex-unlock! *rmt-mutex*) (debug:print-info 12 *default-log-port* "rmt:send-receive, case 6") (rmt:send-receive cmd rid params attemptnum: attemptnum)) ;;DOT CASE7 [label="homehost\nwrite"]; ;;DOT MUTEXLOCK -> CASE7 [label="server not required,\non homehost,\na write,\nhave a server"]; {rank=same "case 7" CASE7}; |
︙ | ︙ | |||
200 201 202 203 204 205 206 | ;;DOT CASE8 -> "rmt:open-qry-close-locally"; ;; on homehost, no server contact made and this is a write, passively start a server ((and (not (remote-force-server runremote)) ;; honor forced use of server, i.e. server NOT required (cdr (remote-hh-dat runremote)) ;; have homehost (not (remote-server-url runremote)) ;; no connection yet (not (member cmd api:read-only-queries))) ;; not a read-only query (debug:print-info 12 *default-log-port* "rmt:send-receive, case 8") | | | > | > | | | 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 | ;;DOT CASE8 -> "rmt:open-qry-close-locally"; ;; on homehost, no server contact made and this is a write, passively start a server ((and (not (remote-force-server runremote)) ;; honor forced use of server, i.e. server NOT required (cdr (remote-hh-dat runremote)) ;; have homehost (not (remote-server-url runremote)) ;; no connection yet (not (member cmd api:read-only-queries))) ;; not a read-only query (debug:print-info 12 *default-log-port* "rmt:send-receive, case 8") (let ((server-info (server:check-if-running *toppath*))) ;; (server:read-dotserver->url *toppath*))) ;; (server:check-if-running *toppath*))) ;; Do NOT want to run server:check-if-running - very expensive to do for every write call (if server-info (begin (remote-server-url-set! runremote (server:record->url server-info)) ;; the string can be consumed by the client setup if needed (remote-server-id-set! runremote (server:record->id server-info))) (if (common:force-server?) (server:start-and-wait *toppath*) (server:kind-run *toppath*))) (remote-force-server-set! runremote (common:force-server?)) (mutex-unlock! *rmt-mutex*) (debug:print-info 12 *default-log-port* "rmt:send-receive, case 8.1") (rmt:open-qry-close-locally cmd 0 params))) ;;DOT CASE9 [label="force server\nnot on homehost"]; ;;DOT MUTEXLOCK -> CASE9 [label="no connection\nand either require server\nor not on homehost"]; {rank=same "case 9" CASE9}; ;;DOT CASE9 -> "start\nserver" -> "rmt:send-receive"; ((or (and (remote-force-server runremote) ;; we are forcing a server and don't yet have a connection to one (not (remote-conndat runremote))) (and (not (cdr (remote-hh-dat runremote))) ;; not on a homehost |
︙ | ︙ | |||
261 262 263 264 265 266 267 268 269 270 271 272 273 274 | ;; problems. However it ;; is needed to deal with ;; attemtped ;; communication to ;; servers that have gone ;; away (http-transport:client-api-send-receive 0 conninfo cmd params) ((commfail)(vector #f "communications fail")) ((exn)(vector #f "other fail" (print-call-chain))))) (else (debug:print 0 *default-log-port* "ERROR: transport " (remote-transport runremote) " not supported") (exit)))) ;; No Title | > | 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 | ;; problems. However it ;; is needed to deal with ;; attemtped ;; communication to ;; servers that have gone ;; away (http-transport:client-api-send-receive 0 conninfo cmd params) ((servermismatch) (vector #f "Server id mismatch" )) ((commfail)(vector #f "communications fail")) ((exn)(vector #f "other fail" (print-call-chain))))) (else (debug:print 0 *default-log-port* "ERROR: transport " (remote-transport runremote) " not supported") (exit)))) ;; No Title |
︙ | ︙ | |||
308 309 310 311 312 313 314 | (http-transport:close-connections area-dat: runremote))) (debug:print-info 13 *default-log-port* "rmt:send-receive, case 9. conninfo=" conninfo " dat=" dat " runremote = " runremote) (mutex-unlock! *rmt-mutex*) (if success ;; success only tells us that the transport was ;; successful, have to examine the data to see if ;; there was a detected issue at the other end (extras-transport-succeded *default-log-port* *rmt-mutex* attemptnum runremote res params rid cmd) | > > | | 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 | (http-transport:close-connections area-dat: runremote))) (debug:print-info 13 *default-log-port* "rmt:send-receive, case 9. conninfo=" conninfo " dat=" dat " runremote = " runremote) (mutex-unlock! *rmt-mutex*) (if success ;; success only tells us that the transport was ;; successful, have to examine the data to see if ;; there was a detected issue at the other end (extras-transport-succeded *default-log-port* *rmt-mutex* attemptnum runremote res params rid cmd) (begin (debug:print-error 0 *default-log-port* " dat=" dat) (extras-transport-failed *default-log-port* *rmt-mutex* attemptnum runremote cmd rid params)) ))) (define (rmt:print-db-stats) (let ((fmtstr "~40a~7-d~9-d~20,2-f")) ;; "~20,2-f" (debug:print 18 *default-log-port* "DB Stats\n========") (debug:print 18 *default-log-port* (format #f "~40a~8a~10a~10a" "Cmd" "Count" "TotTime" "Avg")) (for-each (lambda (cmd) |
︙ | ︙ | |||
659 660 661 662 663 664 665 | (map (lambda (run-id) (rmt:send-receive 'test-get-paths-matching-keynames-target-new run-id (list run-id keynames target res testpatt statepatt statuspatt runname))) run-ids)))) (define (rmt:get-prereqs-not-met run-id waitons ref-test-name ref-item-path #!key (mode '(normal))(itemmaps #f)) (rmt:send-receive 'get-prereqs-not-met run-id (list run-id waitons ref-test-name ref-item-path mode itemmaps))) | | | | | | 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 | (map (lambda (run-id) (rmt:send-receive 'test-get-paths-matching-keynames-target-new run-id (list run-id keynames target res testpatt statepatt statuspatt runname))) run-ids)))) (define (rmt:get-prereqs-not-met run-id waitons ref-test-name ref-item-path #!key (mode '(normal))(itemmaps #f)) (rmt:send-receive 'get-prereqs-not-met run-id (list run-id waitons ref-test-name ref-item-path mode itemmaps))) (define (rmt:get-count-tests-running-for-run-id run-id) (rmt:send-receive 'get-count-tests-running-for-run-id run-id (list run-id))) (define (rmt:get-not-completed-cnt run-id) (rmt:send-receive 'get-not-completed-cnt run-id (list run-id))) ;; Statistical queries (define (rmt:get-count-tests-running run-id) (rmt:send-receive 'get-count-tests-running run-id (list run-id))) (define (rmt:get-count-tests-running-for-testname run-id testname) (rmt:send-receive 'get-count-tests-running-for-testname run-id (list run-id testname))) (define (rmt:get-count-tests-running-in-jobgroup run-id jobgroup) (rmt:send-receive 'get-count-tests-running-in-jobgroup run-id (list run-id jobgroup))) |
︙ | ︙ |
Modified runs.scm from [030b929939] to [a0489a126c].
︙ | ︙ | |||
319 320 321 322 323 324 325 | ;; Take advantage of a good place to exit if running the one-pass methodology (if (and (> (runs:dat-can-run-more-tests-count runsdat) 20) (args:get-arg "-one-pass")) (exit 0)) (if (runs:dat-load-mgmt-function runsdat)((runs:dat-load-mgmt-function runsdat))) | | | 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 | ;; Take advantage of a good place to exit if running the one-pass methodology (if (and (> (runs:dat-can-run-more-tests-count runsdat) 20) (args:get-arg "-one-pass")) (exit 0)) (if (runs:dat-load-mgmt-function runsdat)((runs:dat-load-mgmt-function runsdat))) (let* ((num-running (rmt:get-count-tests-running run-id)) (num-running-in-jobgroup (rmt:get-count-tests-running-in-jobgroup run-id jobgroup)) (job-group-limit (let ((jobg-count (configf:lookup *configdat* "jobgroups" jobgroup))) (if (string? jobg-count) (string->number jobg-count) jobg-count)))) (if (> (+ num-running num-running-in-jobgroup) 0) (runs:inc-can-run-more-tests-count runsdat)) ;; (set! *runs:can-run-more-tests-count* (+ *runs:can-run-more-tests-count* 1))) |
︙ | ︙ | |||
1563 1564 1565 1566 1567 1568 1569 | )) extras) extras) '()))) (waitons (delete-duplicates (append (tests:testqueue-get-waitons test-record) extra-waits) equal?)) (newtal (append tal (list hed))) (regfull (>= (length reg) reglen)) | | | 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 | )) extras) extras) '()))) (waitons (delete-duplicates (append (tests:testqueue-get-waitons test-record) extra-waits) equal?)) (newtal (append tal (list hed))) (regfull (>= (length reg) reglen)) (num-running (rmt:get-count-tests-running-for-run-id run-id)) (testdat (make-runs:testdat hed: hed tal: tal reg: reg reruns: reruns test-record: test-record test-name: test-name |
︙ | ︙ | |||
1829 1830 1831 1832 1833 1834 1835 | (rmt:set-var (conc "lunch-complete-" run-id) "yes") ;; now *if* -run-wait we wait for all tests to be done ;; Now wait for any RUNNING tests to complete (if in run-wait mode) ;; (if (runs:dat-load-mgmt-function runsdat)((runs:dat-load-mgmt-function runsdat))) (thread-sleep! 10) ;; I think there is a race condition here. Let states/statuses settle | | | | | 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 | (rmt:set-var (conc "lunch-complete-" run-id) "yes") ;; now *if* -run-wait we wait for all tests to be done ;; Now wait for any RUNNING tests to complete (if in run-wait mode) ;; (if (runs:dat-load-mgmt-function runsdat)((runs:dat-load-mgmt-function runsdat))) (thread-sleep! 10) ;; I think there is a race condition here. Let states/statuses settle (let wait-loop ((num-running (rmt:get-count-tests-running-for-run-id run-id)) (prev-num-running 0)) ;; (debug:print-info 13 *default-log-port* "num-running=" num-running ", prev-num-running=" prev-num-running) (if (and (or (args:get-arg "-run-wait") (equal? (configf:lookup *configdat* "setup" "run-wait") "yes")) (> num-running 0)) (begin ;; Here we mark any old defunct tests as incomplete. Do this every fifteen minutes ;; (debug:print 0 *default-log-port* "Got here eh! num-running=" num-running " (> num-running 0) " (> num-running 0)) (if (> (current-seconds)(+ last-time-incomplete 900)) (let ((actual-num-running (rmt:get-count-tests-running-for-run-id run-id))) (debug:print-info 0 *default-log-port* "Marking stuck tests as INCOMPLETE while waiting for run " run-id ". Running as pid " (current-process-id) " on " (get-host-name)) (set! last-time-incomplete (current-seconds)) ;; FIXME, this might be causing slow down - use of set! (rmt:find-and-mark-incomplete run-id #f) (debug:print-info 0 *default-log-port* "run-wait specified, waiting on " actual-num-running " tests in RUNNING, REMOTEHOSTSTART or LAUNCHED state at " (time->string (seconds->local-time (current-seconds)))))) ;; (if (runs:dat-load-mgmt-function runsdat)((runs:dat-load-mgmt-function runsdat))) (thread-sleep! 5) ;; (if (>= num-running max-concurrent-jobs) 5 1)) (wait-loop (rmt:get-count-tests-running-for-run-id run-id) num-running)))) ;; LET* ((test-record ;; we get here on "drop through". All done! ;; this is moved to runs:run-testes since this function is getting called twice to ensure everthing is completed. ;; (debug:print-info 0 *default-log-port* "Calling Post Hook") ;; (runs:run-post-hook run-id) (debug:print-info 1 *default-log-port* "All tests launched"))) |
︙ | ︙ |
Modified server.scm from [7b2af2dc7e] to [d9eac964a8].
︙ | ︙ | |||
91 92 93 94 95 96 97 98 99 100 101 102 103 104 | ;; Generate a unique signature for this server (define (server:mk-signature) (message-digest-string (md5-primitive) (with-output-to-string (lambda () (write (list (current-directory) (argv))))))) ;; When using zmq this would send the message back (two step process) ;; with spiffy or rpc this simply returns the return data to be returned ;; (define (server:reply return-addr query-sig success/fail result) (debug:print-info 11 *default-log-port* "server:reply return-addr=" return-addr ", result=" result) | > | 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 | ;; Generate a unique signature for this server (define (server:mk-signature) (message-digest-string (md5-primitive) (with-output-to-string (lambda () (write (list (current-directory) (current-process-id) (argv))))))) ;; When using zmq this would send the message back (two step process) ;; with spiffy or rpc this simply returns the return data to be returned ;; (define (server:reply return-addr query-sig success/fail result) (debug:print-info 11 *default-log-port* "server:reply return-addr=" return-addr ", result=" result) |
︙ | ︙ | |||
161 162 163 164 165 166 167 | (system (conc "nbfake " cmdln)) (unsetenv "TARGETHOST_LOGF") (if (get-environment-variable "TARGETHOST")(unsetenv "TARGETHOST")) (thread-join! log-rotate) (pop-directory))) ;; given a path to a server log return: host port startseconds | > | | | | > > | | > > > | | 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 | (system (conc "nbfake " cmdln)) (unsetenv "TARGETHOST_LOGF") (if (get-environment-variable "TARGETHOST")(unsetenv "TARGETHOST")) (thread-join! log-rotate) (pop-directory))) ;; given a path to a server log return: host port startseconds ;; any changes to number of elements returned by this fuction will dirctly affect server:record->url,server:record->id,server:kill,server:get-num-alive which uses match let (define (server:logf-get-start-info logf) (let ((rx (regexp "^SERVER STARTED: (\\S+):(\\d+) AT ([\\d\\.]+) server-id: (\\S+)"))) ;; SERVER STARTED: host:port AT timesecs server id (handle-exceptions exn (begin (debug:print-info 0 *default-log-port* "failed to get server info from " logf ", exn=" exn) (list #f #f #f #f)) ;; no idea what went wrong, call it a bad server (with-input-from-file logf (lambda () (let loop ((inl (read-line)) (lnum 0)) (if (not (eof-object? inl)) (let ((mlst (string-match rx inl))) (if (not mlst) (if (< lnum 500) ;; give up if more than 500 lines of server log read (loop (read-line)(+ lnum 1)) (begin (debug:print-info 0 *default-log-port* "failed to get server info from first 500 lines of " logf ) (list #f #f #f #f))) (let ((dat (cdr mlst))) (list (car dat) ;; host (string->number (cadr dat)) ;; port (string->number (caddr dat)) (cadr (cddr dat)))))) (begin (debug:print-info 0 *default-log-port* "failed to get server info from " logf " at " (current-seconds)) (list #f #f #f #f))))))))) ;; get a list of servers with all relevant data ;; ( mod-time host port start-time pid ) ;; (define (server:get-list areapath #!key (limit #f)) (let ((fname-rx (regexp "^(|.*/)server-(\\d+)-(\\S+).log$")) (day-seconds (* 24 60 60))) |
︙ | ︙ | |||
215 216 217 218 219 220 221 | '() (let loop ((hed (car server-logs)) (tal (cdr server-logs)) (res '())) (let* ((mod-time (handle-exceptions exn (begin | | | > > > > | | | 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 | '() (let loop ((hed (car server-logs)) (tal (cdr server-logs)) (res '())) (let* ((mod-time (handle-exceptions exn (begin (debug:print 0 *default-log-port* "failed to get modification time on " hed ", exn=" exn) (current-seconds)) ;; 0 (file-modification-time hed))) ;; default to *very* old so log gets ignored if deleted (down-time (- (current-seconds) mod-time)) (serv-dat (if (or (< num-serv-logs 10) (< down-time 900)) ;; day-seconds)) (server:logf-get-start-info hed) '())) ;; don't waste time processing server files not touched in the 15 minutes if there are more than ten servers to look at (serv-rec (cons mod-time serv-dat)) (fmatch (string-match fname-rx hed)) (pid (if fmatch (string->number (list-ref fmatch 2)) #f)) (new-res (if (null? serv-dat) res (cons (append serv-rec (list pid)) res)))) ;; any changes to number of elements in new-res will dirctly affect server:record->url,server:record->id,server:kill,server:get-num-alive which uses match let (if (null? tal) (if (and limit (> (length new-res) limit)) new-res ;; (take new-res limit) <= need intelligent sorting before this will work new-res) (loop (car tal)(cdr tal) new-res))))))))) (define (server:get-num-alive srvlst) (let ((num-alive 0)) (for-each (lambda (server) (handle-exceptions exn (begin (debug:print-info 0 *default-log-port* "failed to get server start-time and/or mod-time from " server ", exn=" exn)) (match-let (((mod-time host port start-time server-id pid) server)) (let* ((uptime (- (current-seconds) mod-time)) (runtime (if start-time (- mod-time start-time) 0))) (if (< uptime 5)(set! num-alive (+ num-alive 1))))))) srvlst) num-alive)) ;; given a list of servers get a list of valid servers, i.e. at least ;; 10 seconds old, has started and is less than 1 hour old and is ;; active (i.e. mod-time < 10 seconds ;; |
︙ | ︙ | |||
303 304 305 306 307 308 309 310 311 | (if (and (list? srvrs) (not (null? srvrs))) (let* ((len (length srvrs)) (idx (random len))) (list-ref srvrs idx)) #f))) (define (server:record->url servr) | > > > > > > > > > > > > > > > > | | | 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 | (if (and (list? srvrs) (not (null? srvrs))) (let* ((len (length srvrs)) (idx (random len))) (list-ref srvrs idx)) #f))) (define (server:record->id servr) (handle-exceptions exn (begin (debug:print-info 0 *default-log-port* "failed to get server id from " servr ", exn=" exn) #f) (match-let (((mod-time host port start-time server-id pid) servr)) (if server-id server-id #f)))) (define (server:record->url servr) (handle-exceptions exn (begin (debug:print-info 0 *default-log-port* "failed to get server url from " servr ", exn=" exn) #f) (match-let (((mod-time host port start-time server-id pid) servr)) (if (and host port) (conc host ":" port) #f)))) (define (server:get-client-signature) ;; BB> why is this proc named "get-"? it returns nothing -- set! has not return value. (if *my-client-signature* *my-client-signature* (let ((sig (server:mk-signature))) (set! *my-client-signature* sig) *my-client-signature*))) |
︙ | ︙ | |||
376 377 378 379 380 381 382 | (common:simple-file-release-lock lock-file))) (hash-table-set! *server-kind-run* areapath (list (+ call-num 1)(current-seconds)))))) ;; this one seems to be the general entry point ;; (define (server:start-and-wait areapath #!key (timeout 60)) (let ((give-up-time (+ (current-seconds) timeout))) | | | | < | > | > > > > > | | | 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 | (common:simple-file-release-lock lock-file))) (hash-table-set! *server-kind-run* areapath (list (+ call-num 1)(current-seconds)))))) ;; this one seems to be the general entry point ;; (define (server:start-and-wait areapath #!key (timeout 60)) (let ((give-up-time (+ (current-seconds) timeout))) (let loop ((server-info (server:check-if-running areapath)) (try-num 0)) (if (or server-info (> (current-seconds) give-up-time)) ;; server-url will be #f if no server available. (server:record->url server-info) (let ((num-ok (length (server:get-best (server:get-list areapath))))) (if (and (> try-num 0) ;; first time through simply wait a little while then try again (< num-ok 1)) ;; if there are no decent candidates for servers then try starting a new one (server:kind-run areapath)) (thread-sleep! 5) (loop (server:check-if-running areapath) (+ try-num 1))))))) (define server:try-running server:run) ;; there is no more per-run servers ;; REMOVE ME. BUG. (define (server:get-num-servers #!key (numservers 2)) (let ((ns (string->number (or (configf:lookup *configdat* "server" "numservers") "notanumber")))) (or ns numservers))) ;; no longer care if multiple servers are started by accident. older servers will drop off in time. ;; (define (server:check-if-running areapath) ;; #!key (numservers "2")) (let* ((ns (server:get-num-servers)) (servers (server:get-best (server:get-list areapath)))) (if (or (and servers (null? servers)) (not servers) (and (list? servers) (< (length servers) (random ns)))) ;; somewhere between 0 and numservers #f (let loop ((hed (car servers)) (tal (cdr servers))) (let ((res (server:check-server hed))) (if res hed (if (null? tal) #f (loop (car tal)(cdr tal))))))))) ;; ping the given server ;; (define (server:check-server server-record) (let* ((server-url (server:record->url server-record)) (server-id (server:record->id server-record)) (res (case *transport-type* ((http)(server:ping server-url server-id)) ;; ((nmsg)(nmsg-transport:ping (tasks:hostinfo-get-interface server) ))) (if res server-url #f))) (define (server:kill servr) (handle-exceptions exn (begin (debug:print-info 0 *default-log-port* "failed to get host and/or port from " servr ", exn=" exn) #f) (match-let (((mod-time hostname port start-time pid) servr)) (tasks:kill-server hostname pid)))) ;; called in megatest.scm, host-port is string hostname:port ;; ;; NOTE: This is NOT called directly from clients as not all transports support a client running ;; in the same process as the server. ;; (define (server:ping host-port-in server-id #!key (do-exit #f)) (let ((host:port (if (not host-port-in) ;; use read-dotserver to find #f ;; (server:check-if-running *toppath*) ;; (if (number? host-port-in) ;; we were handed a server-id ;; (let ((srec (tasks:get-server-by-id (db:delay-if-busy (tasks:open-db)) host-port-in))) ;; ;; (print "srec: " srec " host-port-in: " host-port-in) ;; (if srec ;; (conc (vector-ref srec 3) ":" (vector-ref srec 4)) |
︙ | ︙ | |||
465 466 467 468 469 470 471 | (begin (if host-port-in (debug:print 0 *default-log-port* "ERROR: bad host:port")) (if do-exit (exit 1)) #f) (let* ((iface (car host-port)) (port (cadr host-port)) | | | 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 | (begin (if host-port-in (debug:print 0 *default-log-port* "ERROR: bad host:port")) (if do-exit (exit 1)) #f) (let* ((iface (car host-port)) (port (cadr host-port)) (server-dat (http-transport:client-connect iface port server-id)) (login-res (rmt:login-no-auto-client-setup server-dat))) (if (and (list? login-res) (car login-res)) (begin ;; (print "LOGIN_OK") (if do-exit (exit 0)) #t) |
︙ | ︙ | |||
587 588 589 590 591 592 593 594 595 596 597 | (delete-file* staging-file) (let* ((start-time (current-milliseconds)) (res (system sync-cmd)) (dbbackupfile (conc mtdbfile ".backup")) (res2 (cond ((eq? 0 res ) (if (file-exists? dbbackupfile) (delete-file* dbbackupfile) ) (if (eq? 0 (file-size sync-log)) | > > > | | | 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 | (delete-file* staging-file) (let* ((start-time (current-milliseconds)) (res (system sync-cmd)) (dbbackupfile (conc mtdbfile ".backup")) (res2 (cond ((eq? 0 res ) (handle-exceptions exn #f (if (file-exists? dbbackupfile) (delete-file* dbbackupfile) ) (if (eq? 0 (file-size sync-log)) (delete-file* sync-log)) (system (conc "/bin/mv " staging-file " " mtdbfile)) (set! last-sync-seconds (/ (- (current-milliseconds) start-time) 1000)) (set! off-time (calculate-off-time last-sync-seconds (cond ((and (number? duty-cycle) (> duty-cycle 0) (< duty-cycle 1)) duty-cycle) (else (debug:print 0 *default-log-port* "WARNING: ["(common:human-time)"] server.sync-duty-cycle is invalid. Should be a number between 0 and 1, but "duty-cycle" was specified. Using default value: "default-duty-cycle) default-duty-cycle)))) (debug:print 1 *default-log-port* "INFO: ["(common:human-time)"] pid="(current-process-id)" SYNC took "last-sync-seconds" sec") (debug:print 1 *default-log-port* "INFO: ["(common:human-time)"] pid="(current-process-id)" SYNC took "last-sync-seconds" sec ; with duty-cycle of "duty-cycle" off time is now "off-time) 'sync-completed)) (else (system (conc "/bin/cp "sync-log" "sync-log".fail")) (debug:print 0 *default-log-port* "ERROR: ["(common:human-time)"] Sync failed. See log at "sync-log".fail") (if (file-exists? (conc mtdbfile ".backup")) (system (conc "/bin/cp "mtdbfile ".backup " mtdbfile))) #f)))) (common:simple-file-release-lock lockfile) |
︙ | ︙ |
Modified sretrieve.scm from [e7efdf8d00] to [c73e7e987b].
︙ | ︙ | |||
635 636 637 638 639 640 641 | (let* ((parent-dir target-path) (last-dir-name (if (pathname-extension target-path) (conc(pathname-file target-path) "." (pathname-extension target-path)) (pathname-file target-path))) (curr-dir (current-directory)) (start-dir (conc (current-directory) "/" last-dir-name)) (execlude (make-exclude-pattern (string-split restrictions ","))) | | | | | | 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 | (let* ((parent-dir target-path) (last-dir-name (if (pathname-extension target-path) (conc(pathname-file target-path) "." (pathname-extension target-path)) (pathname-file target-path))) (curr-dir (current-directory)) (start-dir (conc (current-directory) "/" last-dir-name)) (execlude (make-exclude-pattern (string-split restrictions ","))) (tmpfile (conc "/tmp/my-pipe-" (current-process-id)))) (if (file-exists? start-dir) (begin (sauth:print-error (conclast-dir-name " already exist in your work dir.")) (sauth:print-error "Nothing has been retrieved!! ")) (begin ; (sretrieve:do-as-calling-user ; (lambda () ; (print tmpfile) ;(if (not (file-exists? (conc "/tmp/" (current-user-name)))) ; (create-directory (conc "/tmp/" (current-user-name)) #t)) (change-directory parent-dir) (create-fifo tmpfile) (process-fork (lambda() (sleep 1) (with-output-to-file tmpfile (lambda () |
︙ | ︙ |
Modified tasks.scm from [b621e9649f] to [a73c5b318e].
︙ | ︙ | |||
753 754 755 756 757 758 759 | (state (db:get-value-by-header row header "state")) (status (db:get-value-by-header row header "status")) (owner (db:get-value-by-header row header "owner")) (event-time (db:get-value-by-header row header "event_time")) (comment (db:get-value-by-header row header "comment")) (fail-count (db:get-value-by-header row header "fail_count")) (pass-count (db:get-value-by-header row header "pass_count")) | | | | | > | | | | | 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 | (state (db:get-value-by-header row header "state")) (status (db:get-value-by-header row header "status")) (owner (db:get-value-by-header row header "owner")) (event-time (db:get-value-by-header row header "event_time")) (comment (db:get-value-by-header row header "comment")) (fail-count (db:get-value-by-header row header "fail_count")) (pass-count (db:get-value-by-header row header "pass_count")) (db-contour (db:get-value-by-header row header "contour")) (contour (if (args:get-arg "-prepend-contour") (if (and db-contour (not (equal? db-contour "")) (string? db-contour )) (begin (debug:print-info 10 *default-log-port* "db-contour" db-contour) db-contour) (args:get-arg "-contour")))) (run-tag (if (args:get-arg "-run-tag") (args:get-arg "-run-tag") "")) (last-update (db:get-value-by-header row header "last_update")) (keytarg (if (or (args:get-arg "-prepend-contour") (args:get-arg "-prefix-target")) (conc "MT_CONTOUR/MT_AREA/" (string-intersperse (rmt:get-keys) "/")) (string-intersperse (rmt:get-keys) "/"))) ;; e.g. version/iteration/platform (base-target (rmt:get-target run-id)) (target (if (or (args:get-arg "-prepend-contour") (args:get-arg "-prefix-target")) (conc (or (args:get-arg "-prefix-target") (conc contour "/" (common:get-area-name) "/")) base-target) base-target)) ;; e.g. v1.63/a3e1/ubuntu (spec-id (pgdb:get-ttype dbh keytarg)) (publish-time (if (args:get-arg "-cp-eventtime-to-publishtime") event-time (current-seconds))) (new-run-id (if (and run-name base-target) (pgdb:get-run-id dbh spec-id target run-name area-id) #f))) (if new-run-id (begin ;; let ((run-record (pgdb:get-run-info dbh new-run-id)) (hash-table-set! runs-ht run-id new-run-id) ;; ensure key fields are up to date ;; if last_update == pgdb_last_update do not update smallest-last-update-time (let* ((pgdb-last-update (pgdb:get-run-last-update dbh new-run-id)) (smallest-time (hash-table-ref/default smallest-last-update-time "smallest-time" #f))) (if (and (> last-update pgdb-last-update) (or (not smallest-time) (< last-update smallest-time))) (hash-table-set! smallest-last-update-time "smallest-time" last-update))) (pgdb:refresh-run-info dbh new-run-id state status owner event-time comment fail-count pass-count area-id last-update publish-time) (debug:print-info 4 *default-log-port* "Working on run-id " run-id " pgdb-id " new-run-id ) (if (not (equal? run-tag "")) (task:add-run-tag dbh new-run-id run-tag)) new-run-id) (if (or (not state) (equal? state "deleted")) (begin (debug:print-info 1 *default-log-port* "Warning: Run with id " run-id " was created after previous sync and deleted before the sync") #f) (if (handle-exceptions exn (begin (print-call-chain) (print ((condition-property-accessor 'exn 'message) exn)) #f) |
︙ | ︙ | |||
858 859 860 861 862 863 864 | #f))) (if step-id (begin (if pgdb-test-id (begin (if pgdb-step-id (begin | | | | 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 | #f))) (if step-id (begin (if pgdb-test-id (begin (if pgdb-step-id (begin (debug:print-info 4 *default-log-port* "Updating existing test-step with test-id: " test-id " and step-id " step-id " pgdb test id: " pgdb-test-id " pgdb step id " pgdb-step-id ) (let* ((pgdb-last-update (pgdb:get-test-step-last-update dbh pgdb-step-id))) (if (and (> last-update pgdb-last-update) (or (not smallest-time) (< last-update smallest-time))) (hash-table-set! smallest-last-update-time "smallest-time" last-update))) (pgdb:update-test-step dbh pgdb-step-id pgdb-test-id stepname state status event_time comment logfile last-update)) (begin (debug:print-info 4 *default-log-port* "Inserting test-step with test-id: " test-id " and step-id " step-id " pgdb test id: " pgdb-test-id) (if (or (not smallest-time) (< last-update smallest-time)) (hash-table-set! smallest-last-update-time "smallest-time" last-update)) (pgdb:insert-test-step dbh pgdb-test-id stepname state status event_time comment logfile last-update ) (set! pgdb-step-id (pgdb:get-test-step-id dbh pgdb-test-id stepname state)))) (hash-table-set! step-ht step-id pgdb-step-id )) (debug:print-info 1 *default-log-port* "Error: Test not cashed"))) (debug:print-info 1 *default-log-port* "Error: Could not get test step info for step id " test-step-id )))) ;; this is a wierd senario need to debug |
︙ | ︙ | |||
904 905 906 907 908 909 910 | #f))) (if data-id (begin (if pgdb-test-id (begin (if pgdb-data-id (begin | | | | 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 | #f))) (if data-id (begin (if pgdb-test-id (begin (if pgdb-data-id (begin (debug:print-info 4 *default-log-port* "Updating existing test-data with test-id: " test-id " and data-id " data-id " pgdb test id: " pgdb-test-id " pgdb data id " pgdb-data-id) (let* ((pgdb-last-update (pgdb:get-test-data-last-update dbh pgdb-data-id))) (if (and (> last-update pgdb-last-update) (or (not smallest-time) (< last-update smallest-time))) (hash-table-set! smallest-last-update-time "smallest-time" last-update))) (pgdb:update-test-data dbh pgdb-data-id pgdb-test-id category variable value expected tol units comment status type last-update)) (begin (debug:print-info 4 *default-log-port* "Inserting test-data with test-id: " test-id " and data-id " data-id " pgdb test id: " pgdb-test-id) (if (handle-exceptions exn (begin (print-call-chain) (print ((condition-property-accessor 'exn 'message) exn)) #f) (pgdb:insert-test-data dbh pgdb-test-id category variable value expected tol units comment status type last-update)) |
︙ | ︙ | |||
968 969 970 971 972 973 974 975 976 977 978 | (begin ;(print pgdb-run-id) (pgdb:get-test-id dbh pgdb-run-id test-name item-path)) #f))) ;; "id" "run_id" "testname" "state" "status" "event_time" ;; "host" "cpuload" "diskfree" "uname" "rundir" "item_path" ;; "run_duration" "final_logf" "comment" "shortdir" "attemptnum" "archived" (if pgdb-run-id (begin (if pgdb-test-id ;; have a record (begin ;; let ((key-name (conc run-id "/" test-name "/" item-path))) | > > | | | 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 | (begin ;(print pgdb-run-id) (pgdb:get-test-id dbh pgdb-run-id test-name item-path)) #f))) ;; "id" "run_id" "testname" "state" "status" "event_time" ;; "host" "cpuload" "diskfree" "uname" "rundir" "item_path" ;; "run_duration" "final_logf" "comment" "shortdir" "attemptnum" "archived" (if (or (not item-path) (string-null? item-path)) (debug:print-info 0 *default-log-port* "Working on Run id : " run-id "and test name : " test-name)) (if pgdb-run-id (begin (if pgdb-test-id ;; have a record (begin ;; let ((key-name (conc run-id "/" test-name "/" item-path))) (debug:print-info 4 *default-log-port* "Updating existing test with run-id: " run-id " and test-id: " test-id " pgdb run id: " pgdb-run-id " pgdb-test-id " pgdb-test-id) (let* ((pgdb-last-update (pgdb:get-test-last-update dbh pgdb-test-id))) (if (and (> last-update pgdb-last-update) (or (not smallest-time) (< last-update smallest-time))) ;;if last-update is same as pgdb-last-update then it is safe to assume the records are identical and we can use a larger last update time. (hash-table-set! smallest-last-update-time "smallest-time" last-update))) (pgdb:update-test dbh pgdb-test-id pgdb-run-id test-name item-path state status host cpuload diskfree uname run-dir log-file run-duration comment event-time archived last-update pid)) (begin (debug:print-info 4 *default-log-port* "Inserting test with run-id: " run-id " and test-id: " test-id " pgdb run id: " pgdb-run-id) (pgdb:insert-test dbh pgdb-run-id test-name item-path state status host cpuload diskfree uname run-dir log-file run-duration comment event-time archived last-update pid) (if (or (not smallest-time) (< last-update smallest-time)) (hash-table-set! smallest-last-update-time "smallest-time" last-update)) (set! pgdb-test-id (pgdb:get-test-id dbh pgdb-run-id test-name item-path)))) (hash-table-set! test-ht test-id pgdb-test-id)) (debug:print-info 1 *default-log-port* "WARNING: Skipping run with run-id:" run-id ". This run was created after privious sync and removed before this sync.")))) test-ids))) |
︙ | ︙ | |||
1011 1012 1013 1014 1015 1016 1017 | #f) (if (not (pgdb:is-area-taged-with-a-tag dbh (vector-ref tag-info 0) (vector-ref area-info 0))) (pgdb:insert-area-tag dbh (vector-ref tag-info 0) (vector-ref area-info 0)))))) (define (tasks:sync-run-data dbh cached-info run-ids area-info smallest-last-update-time) (for-each (lambda (run-id) | | | 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 | #f) (if (not (pgdb:is-area-taged-with-a-tag dbh (vector-ref tag-info 0) (vector-ref area-info 0))) (pgdb:insert-area-tag dbh (vector-ref tag-info 0) (vector-ref area-info 0)))))) (define (tasks:sync-run-data dbh cached-info run-ids area-info smallest-last-update-time) (for-each (lambda (run-id) (debug:print-info 4 *default-log-port* "Check if run with " run-id " needs to be synced" ) (tasks:run-id->mtpg-run-id dbh cached-info run-id area-info smallest-last-update-time)) run-ids)) ;; get runs changed since last sync ;; (define (tasks:sync-test-data dbh cached-info area-info) ;; (let* (( |
︙ | ︙ | |||
1080 1081 1082 1083 1084 1085 1086 | (debug:print-info 0 *default-log-port* "syncing tests") (tasks:sync-tests-data dbh cached-info test-ids area-info smallest-last-update-time) (debug:print-info 0 *default-log-port* "syncing test steps") (tasks:sync-test-steps dbh cached-info test-step-ids smallest-last-update-time) (debug:print-info 0 *default-log-port* "syncing test data") (tasks:sync-test-gen-data dbh cached-info test-data-ids smallest-last-update-time) (print "----------done---------------"))) | | | 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 | (debug:print-info 0 *default-log-port* "syncing tests") (tasks:sync-tests-data dbh cached-info test-ids area-info smallest-last-update-time) (debug:print-info 0 *default-log-port* "syncing test steps") (tasks:sync-test-steps dbh cached-info test-step-ids smallest-last-update-time) (debug:print-info 0 *default-log-port* "syncing test data") (tasks:sync-test-gen-data dbh cached-info test-data-ids smallest-last-update-time) (print "----------done---------------"))) (let* ((smallest-time (hash-table-ref/default smallest-last-update-time "smallest-time" (current-seconds)))) (debug:print-info 0 "smallest-time :" smallest-time " last-sync-time " last-sync-time) (if (not (and target run-name)) (if (or (and smallest-time (> smallest-time last-sync-time)) (and smallest-time (eq? last-sync-time 0))) (pgdb:write-sync-time dbh area-info smallest-time))))) ;;this needs to be changed (if (tasks:set-area dbh configdat) (tasks:sync-to-postgres configdat dest) (begin (debug:print 0 *default-log-port* "ERROR: unable to create an area record") #f))))) |