Megatest

(declare (uses common))
(declare (uses configf))
(declare (uses db))
(declare (uses ezsteps))

(declare (uses commonmod))
(import commonmod)







(include "common_records.scm")
(include "key_records.scm")
(include "db_records.scm")
(include "megatest-fossil-hash.scm")

;;======================================================================

    ;; (tests:set-full-meta-info test-id run-id (calc-minutes) work-area)
    (tests:set-full-meta-info #f test-id run-id (calc-minutes) work-area 10)

    (let loop ((minutes   (calc-minutes))
	       (cpu-load  (alist-ref 'adj-core-load (common:get-normalized-cpu-load #f)))
	       (disk-free (get-df (current-directory)))
               (last-sync (current-seconds)))
      (common:telemetry-log "zombie" (conc "launch:monitor-job - top of loop encountered at "(current-seconds)" with last-sync="last-sync))
      (let* ((over-time     (> (current-seconds) (+ last-sync update-period)))
             (new-cpu-load  (let* ((load  (alist-ref 'adj-core-load (common:get-normalized-cpu-load #f)))
                                   (delta (abs (- load cpu-load))))
                              (if (> delta 0.1) ;; don't bother updating with small changes
                                  load
                                  #f)))
             (new-disk-free (let* ((df    (if over-time ;; only get df every 30 seconds

             (do-sync       (or new-cpu-load new-disk-free over-time))

             (test-info   (rmt:get-test-info-by-id run-id test-id))
             (state       (db:test-get-state test-info))
             (status      (db:test-get-status test-info))
             (kill-reason  "no kill reason specified")
             (kill-job?    #f))
        (common:telemetry-log "zombie" (conc "launch:monitor-job - decision time encountered at "(current-seconds)" with last-sync="last-sync" do-sync="do-sync" over-time="over-time" update-period="update-period))
        (cond
         ((test-get-kill-request run-id test-id)
          (set! kill-reason "KILLING TEST since received kill request (KILLREQ)")
          (set! kill-job? #t))
         ((and runtlim (> (- (current-seconds) start-seconds) runtlim))
          (set! kill-reason (conc "KILLING TEST DUE TO TIME LIMIT EXCEEDED! Runtime=" (- (current-seconds) start-seconds) " seconds, limit=" runtlim))
          (set! kill-job? #t))
         ((equal? status "DEAD")
          (tests:update-central-meta-info run-id test-id new-cpu-load new-disk-free (calc-minutes) #f #f)
          (rmt:set-state-status-and-roll-up-items run-id test-id 'foo "RUNNING" "n/a" "was marked dead; really still running.")
          ;;(set! kill-reason "KILLING TEST because it was marked as DEAD by launch:handle-zombie-tests (might indicate really overloaded server or else overzealous setup.deadtime)") ;; MARK RUNNING
          (set! kill-job? #f)))

        (debug:print 4 *default-log-port* "cpu: " new-cpu-load " disk: " new-disk-free " last-sync: " last-sync " do-sync: " do-sync)
        (launch:handle-zombie-tests run-id)
        (when do-sync
          ;;(with-output-to-file (conc (getenv "MT_TEST_RUN_DIR") "/last-loadinfo.log" #:append)
          ;;  (lambda () (pp (list (current-seconds) new-cpu-load new-disk-free (calc-minutes)))))
          (common:telemetry-log "zombie" (conc  "launch:monitor-job - dosync started at "(current-seconds)))
          (tests:update-central-meta-info run-id test-id new-cpu-load new-disk-free (calc-minutes) #f #f)
          (common:telemetry-log "zombie" (conc "launch:monitor-job - dosync finished at "(current-seconds))))
        
	(if kill-job? 
	    (begin
              (debug:print-info 0 *default-log-port* "proceeding to kill test: "kill-reason)
	      (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

			     (tal    (cdr not-completed-tests)))
		       (let* ((test-name (vector-ref running-test 2))
                 (item-path (vector-ref running-test 11)))
			       	(debug:print 0 *default-log-port* "test " test-name "/" item-path " not completed")
              (if (not (null? tal))
				  (loop (car tal) (cdr tal)))))))))))        
 
(define (launch:is-test-alive host pid)
  (if (and host pid (not (equal? host "n/a")))
      (let* ((cmd (conc "ssh " host " pstree -A " pid))
	     (output (with-input-from-pipe cmd read-lines)))
	(debug:print 2 *default-log-port* "Running " cmd " received " output)
	(if (eq? (length output) 0)
	   #f
	   #t))
      #t))
 
(define (launch:kill-tests-if-dead run-id)
  (let* ((running-tests (rmt:get-tests-for-run run-id "%" `("RUNNING" "LAUNCHED" "REMOTEHOSTSTART") `() #f #f #f #f #f #f #f #f)))
       (let loop ((running-test (car running-tests))
			     (tal    (cdr running-tests))
			     (kill-cnt 0))
		       (let* ((test-name (vector-ref running-test 2))
                 (item-path (vector-ref running-test 11))