328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
|
(begin
(mutex-lock! m)
;; NOTE: The pid can change as different steps are run. Do we need handshaking between this
;; section and the runit section? Or add a loop that tries three times with a 1/4 second
;; between tries?
(let* ((pid (vector-ref exit-info 0)))
(if (number? pid)
(process-signal pid signal/kill)
;; (begin
;; (debug:print 0 "WARNING: Request received to kill job (attempt # " kill-tries ")")
;; (let ((processes (cmd-run->list (conc "pgrep -l -P " pid))))
;; (for-each
;; (lambda (p)
;; (let* ((parts (string-split p))
;; (p-id (if (> (length parts) 0)
;; (string->number (car parts))
;; #f)))
;; (if p-id
;; (begin
;; (debug:print 0 "Killing " (cadr parts) "; kill -9 " p-id)
;; (system (conc "kill -9 " p-id))))))
;; (car processes))
;; (system (conc "kill -9 -" pid))))
(begin
(debug:print 0 "WARNING: Request received to kill job but problem with process, attempting to kill manager process")
(tests:test-set-status! run-id test-id "KILLED" "FAIL"
(args:get-arg "-m") #f)
(exit 1) ;; IS THIS NECESSARY OR WISE???
)))
(set! kill-tries (+ 1 kill-tries))
(mutex-unlock! m)))
(if keep-going
(begin
(thread-sleep! 3) ;; (+ 3 (random 6))) ;; add some jitter to the call home time to spread out the db accesses
|
>
>
>
|
|
|
|
|
|
|
|
|
|
|
|
|
>
|
|
|
>
|
|
|
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
|
(begin
(mutex-lock! m)
;; NOTE: The pid can change as different steps are run. Do we need handshaking between this
;; section and the runit section? Or add a loop that tries three times with a 1/4 second
;; between tries?
(let* ((pid (vector-ref exit-info 0)))
(if (number? pid)
(handle-exceptions
exn
(debug:print-info 0 "Unable to kill process with pid " pid ", possibly already killed.")
;;(process-signal pid signal/kill))
(begin
(debug:print 0 "WARNING: Request received to kill job (attempt # " kill-tries ")")
(let ((processes (cmd-run->list (conc "pgrep -l -P " pid))))
(for-each
(lambda (p)
(let* ((parts (string-split p))
(p-id (if (> (length parts) 0)
(string->number (car parts))
#f)))
(if p-id
(begin
(debug:print 0 "Killing " (cadr parts) "; kill -9 " p-id)
;; (process-signal pid signal/kill))))) ;;
(system (conc "kill -9 " p-id))))))
(car processes)))
(system (conc "kill -9 -" pid))
(tests:test-set-status! test-id "KILLED" "FAIL" (args:get-arg "-m") #f)))
(begin
(debug:print 0 "WARNING: Request received to kill job but problem with process, attempting to kill manager process")
;; (tests:test-set-status! run-id test-id "KILLED" "FAIL"
(tests:test-set-status! run-id test-id "KILLED" "FAIL" (args:get-arg "-m") #f)
(exit 1) ;; IS THIS NECESSARY OR WISE???
)))
(set! kill-tries (+ 1 kill-tries))
(mutex-unlock! m)))
(if keep-going
(begin
(thread-sleep! 3) ;; (+ 3 (random 6))) ;; add some jitter to the call home time to spread out the db accesses
|