resolve mutual dependency in flink via dataflow

This commit is contained in:
Sam Caldwell 2019-05-24 11:42:21 -04:00
parent c37c060dc9
commit 98a779bdc1
1 changed files with 31 additions and 27 deletions

View File

@ -369,22 +369,8 @@ The JobManager then performs the job and, when finished, asserts (job-finished I
(match-define (task $this-id $desc) t)
(log "JM begins on task ~a" this-id)
(define ( (ρ2) (select-a-task-manager [assign-task : (proc ID -> ★/t
#:roles (ρ2))]))
(start-facet this-facet
(begin/dataflow
(define mngr?
(for/first ([(id slots) (ref task-managers)]
#:when (positive? (- slots (hash-ref/failure (ref requests-in-flight) id 0))))
id))
(match mngr?
[(some $mngr)
(take-slot! mngr)
(stop this-facet
(assign-task mngr))]
[none
#f])
#f)))
(define not-a-real-task-manager (gensym 'FAKE))
(field [task-mngr ID not-a-real-task-manager])
;; ID -> ...
(define (assign-task [mngr : ID])
@ -392,16 +378,17 @@ The JobManager then performs the job and, when finished, asserts (job-finished I
(assert (task-assignment mngr job-id t))
(on (retracted (task-manager mngr discard))
;; our task manager has crashed
(stop this-facet #;(select-a-task-manager)))
(stop this-facet
(set! task-mngr not-a-real-task-manager)))
(on start
;; N.B. when this line was here, and not after `(when mngr ...)` above,
;; things didn't work. I think that due to script scheduling, all ready
;; tasks were being assigned to the manager
#;(take-slot! mngr)
(start-facet take-slot
(on (asserted (task-state mngr job-id this-id _))
(stop take-slot
(received-answer! mngr)))))
;; N.B. when this line was here, and not after `(when mngr ...)` above,
;; things didn't work. I think that due to script scheduling, all ready
;; tasks were being assigned to the manager
#;(take-slot! mngr)
(start-facet take-slot
(on (asserted (task-state mngr job-id this-id _))
(stop take-slot
(received-answer! mngr)))))
(on (asserted (task-state mngr job-id this-id $status))
(match status
[ACCEPTED #f]
@ -411,12 +398,29 @@ The JobManager then performs the job and, when finished, asserts (job-finished I
;; don't think we need a release-slot! here, because if we've heard back from a task manager,
;; they should have told us a different slot count since we tried to give them work
(log "JM overloaded manager ~a with task ~a" mngr this-id)
(stop this-facet #;(select-a-task-manager))]
(stop this-facet
(set! task-mngr not-a-real-task-manager))]
[(finished $results)
(log "JM receives the results of task ~a" this-id)
(stop perform (k this-id results))]))))
(on start (select-a-task-manager assign-task))))
(define (select-a-task-manager)
(start-facet this-facet
(begin/dataflow
(when (equal? (ref task-mngr) not-a-real-task-manager)
(define mngr?
(for/first ([(id slots) (ref task-managers)]
#:when (positive? (- slots (hash-ref/failure (ref requests-in-flight) id 0))))
id))
(match mngr?
[(some $mngr)
(take-slot! mngr)
(set! task-mngr mngr)
(assign-task mngr)]
[none
#f])))))
(on start (select-a-task-manager))))
(begin/dataflow
(define slots (slots-available))