@@ -211,11 +211,13 @@ (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)) + ;; (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 @@ -233,11 +235,11 @@ (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)) + #;(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)) @@ -465,10 +467,13 @@ (debug:print 0 *default-log-port* "ERROR: failed to find a record for test-id " test-id ", exiting.") (exit)))) (test-pid (db:test-get-process_id test-info))) (cond ;; -mrw- I'm removing KILLREQ from this list so that a test in KILLREQ state is treated as a "do not run" flag. + ;;((or (member (db:test-get-state test-info) '("INCOMPLETE" "KILLED" "UNKNOWN" "STUCK")) ;; prior run of this test didn't complete, go ahead and try to rerun + ;; (and (equal? (db:test-get-state test-info) "COMPLETED") ;; completed/abort => rerun if asked + ;; (member (db:test-get-status test-info) '("ABORT")))) ((member (db:test-get-state test-info) '("INCOMPLETE" "KILLED" "UNKNOWN" "STUCK")) ;; prior run of this test didn't complete, go ahead and try to rerun (debug:print 0 *default-log-port* "INFO: test is INCOMPLETE or KILLED, treat this execute call as a rerun request") ;; (tests:test-force-state-status! run-id test-id "REMOTEHOSTSTART" "n/a") (rmt:general-call 'set-test-start-time #f test-id)