371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
|
(configf:lookup *configdat* "server" "required"))
;; (maxqry (cdr (rmt:get-max-query-average run-id)))
;; (threshold (string->number (or (configf:lookup *configdat* "server" "server-query-threshold") "10"))))
;; (cond
;; (forced
;; (if (common:low-noise-print 60 run-id "server required is set")
;; (debug:print-info 0 #f "Server required is set, starting server for run-id " run-id "."))
;; #t)
;; ((> maxqry threshold)
;; (if (common:low-noise-print 60 run-id "Max query time execeeded")
;; (debug:print-info 0 #f "Max avg query time of " maxqry "ms exceeds limit of " threshold "ms, server needed for run-id " run-id "."))
;; #t)
;; (else
;; #f))))
;; try to start a server and wait for it to be available
;;
(define (tasks:start-and-wait-for-server tdbdat run-id delay-max-tries)
|
|
|
|
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
|
(configf:lookup *configdat* "server" "required"))
;; (maxqry (cdr (rmt:get-max-query-average run-id)))
;; (threshold (string->number (or (configf:lookup *configdat* "server" "server-query-threshold") "10"))))
;; (cond
;; (forced
;; (if (common:low-noise-print 60 run-id "server required is set")
;; (debug:print-info 0 *default-log-port* "Server required is set, starting server for run-id " run-id "."))
;; #t)
;; ((> maxqry threshold)
;; (if (common:low-noise-print 60 run-id "Max query time execeeded")
;; (debug:print-info 0 *default-log-port* "Max avg query time of " maxqry "ms exceeds limit of " threshold "ms, server needed for run-id " run-id "."))
;; #t)
;; (else
;; #f))))
;; try to start a server and wait for it to be available
;;
(define (tasks:start-and-wait-for-server tdbdat run-id delay-max-tries)
|
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
|
FROM servers WHERE run_id=? AND state NOT LIKE 'defunct%' ORDER BY start_time DESC;"
run-id)
(reverse res)))
;; no elegance here ...
;;
(define (tasks:kill-server hostname pid)
(debug:print-info 0 #f "Attempting to kill server process " pid " on host " hostname)
(setenv "TARGETHOST" hostname)
(setenv "TARGETHOST_LOGF" "server-kills.log")
(system (conc "nbfake kill " pid))
(unsetenv "TARGETHOST_LOGF")
(unsetenv "TARGETHOST"))
;; look up a server by run-id and send it a kill, also delete the record for that server
;;
(define (tasks:kill-server-run-id run-id #!key (tag "default"))
(let* ((tdbdat (tasks:open-db))
(sdat (tasks:get-server (db:delay-if-busy tdbdat) run-id)))
(if sdat
(let ((hostname (vector-ref sdat 6))
(pid (vector-ref sdat 5))
(server-id (vector-ref sdat 0)))
(tasks:server-set-state! (db:delay-if-busy tdbdat) server-id "killed")
(debug:print-info 0 #f "Killing server " server-id " for run-id " run-id " on host " hostname " with pid " pid)
(tasks:kill-server hostname pid)
(tasks:server-delete-record (db:delay-if-busy tdbdat) server-id tag) )
(debug:print-info 0 #f "No server found for run-id " run-id ", nothing to kill"))
;; (sqlite3:finalize! tdb)
))
;;======================================================================
;; M O N I T O R S
;;======================================================================
|
|
|
|
|
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
|
FROM servers WHERE run_id=? AND state NOT LIKE 'defunct%' ORDER BY start_time DESC;"
run-id)
(reverse res)))
;; no elegance here ...
;;
(define (tasks:kill-server hostname pid)
(debug:print-info 0 *default-log-port* "Attempting to kill server process " pid " on host " hostname)
(setenv "TARGETHOST" hostname)
(setenv "TARGETHOST_LOGF" "server-kills.log")
(system (conc "nbfake kill " pid))
(unsetenv "TARGETHOST_LOGF")
(unsetenv "TARGETHOST"))
;; look up a server by run-id and send it a kill, also delete the record for that server
;;
(define (tasks:kill-server-run-id run-id #!key (tag "default"))
(let* ((tdbdat (tasks:open-db))
(sdat (tasks:get-server (db:delay-if-busy tdbdat) run-id)))
(if sdat
(let ((hostname (vector-ref sdat 6))
(pid (vector-ref sdat 5))
(server-id (vector-ref sdat 0)))
(tasks:server-set-state! (db:delay-if-busy tdbdat) server-id "killed")
(debug:print-info 0 *default-log-port* "Killing server " server-id " for run-id " run-id " on host " hostname " with pid " pid)
(tasks:kill-server hostname pid)
(tasks:server-delete-record (db:delay-if-busy tdbdat) server-id tag) )
(debug:print-info 0 *default-log-port* "No server found for run-id " run-id ", nothing to kill"))
;; (sqlite3:finalize! tdb)
))
;;======================================================================
;; M O N I T O R S
;;======================================================================
|
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
|
"SELECT count(id) FROM monitors WHERE last_update < (strftime('%s','now') - 300) AND username=?;"
(car (user-information (current-user-id))))
res))
;;
(define (tasks:start-monitor db mdb)
(if (> (tasks:get-num-alive-monitors mdb) 2) ;; have two running, no need for more
(debug:print-info 1 #f "Not starting monitor, already have more than two running")
(let* ((megatestdb (conc *toppath* "/megatest.db"))
(monitordbf (conc (db:dbfile-path #f) "/monitor.db"))
(last-db-update 0)) ;; (file-modification-time megatestdb)))
(task:register-monitor mdb)
(let loop ((count 0)
(next-touch 0)) ;; next-touch is the time where we need to update last_update
;; if the db has been modified we'd best look at the task queue
|
|
|
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
|
"SELECT count(id) FROM monitors WHERE last_update < (strftime('%s','now') - 300) AND username=?;"
(car (user-information (current-user-id))))
res))
;;
(define (tasks:start-monitor db mdb)
(if (> (tasks:get-num-alive-monitors mdb) 2) ;; have two running, no need for more
(debug:print-info 1 *default-log-port* "Not starting monitor, already have more than two running")
(let* ((megatestdb (conc *toppath* "/megatest.db"))
(monitordbf (conc (db:dbfile-path #f) "/monitor.db"))
(last-db-update 0)) ;; (file-modification-time megatestdb)))
(task:register-monitor mdb)
(let loop ((count 0)
(next-touch 0)) ;; next-touch is the time where we need to update last_update
;; if the db has been modified we'd best look at the task queue
|