Megatest

Diff
Login

Differences From Artifact [513fa82301]:

To Artifact [d409eb1269]:


161
162
163
164
165
166
167
168
169
170
171




172
173

174
175
176

177
178
179
180
181
182
183


184
185
186
187
188
189
190
161
162
163
164
165
166
167


168

169
170
171
172
173

174
175
176

177
178
179
180




181
182
183
184
185
186
187
188
189







-
-

-
+
+
+
+

-
+


-
+



-
-
-
-
+
+








(define (set-item-env-vars itemdat)
  (for-each (lambda (item)
	      (debug:print 2 "setenv " (car item) " " (cadr item))
	      (setenv (car item) (cadr item)))
	    itemdat))

(define *last-num-running-tests* 0)

;; Every time can-run-more-tests is called increment the delay
;; if the cou
;;
;; NOTE: We run this server-side!! Do not use this global except in the runs:can-run-more-tests routine
;;
(define *last-num-running-tests* 0)
(define *runs:can-run-more-tests-count* 0)
(define (runs:shrink-can-run-more-tests-count)
(define (runs:shrink-can-run-more-tests-count db) ;; the db is just so we can use cdb:remote-run
  (set! *runs:can-run-more-tests-count* 0)) ;; (/ *runs:can-run-more-tests-count* 2)))

(define (runs:can-run-more-tests test-record max-concurrent-jobs)
(define (runs:can-run-more-tests db jobgroup max-concurrent-jobs)
  (thread-sleep! (cond
		  ((> *runs:can-run-more-tests-count* 20) 2);; obviously haven't had any work to do for a while
		  (else 0)))
  (let* ((tconfig                 (tests:testqueue-get-testconfig test-record))
	 (jobgroup                (config-lookup tconfig "requirements" "jobgroup"))
	 (num-running             (cdb:remote-run db:get-count-tests-running #f))
	 (num-running-in-jobgroup (cdb:remote-run db:get-count-tests-running-in-jobgroup #f jobgroup))
  (let* ((num-running             (db:get-count-tests-running db))
	 (num-running-in-jobgroup (db:get-count-tests-running-in-jobgroup db jobgroup))
	 (job-group-limit         (config-lookup *configdat* "jobgroups" jobgroup)))
    (if (> (+ num-running num-running-in-jobgroup) 0)
	(set! *runs:can-run-more-tests-count* (+ *runs:can-run-more-tests-count* 1)))
    (if (not (eq? *last-num-running-tests* num-running))
	(begin
	  (debug:print 2 "max-concurrent-jobs: " max-concurrent-jobs ", num-running: " num-running)
	  (set! *last-num-running-tests* num-running)))
612
613
614
615
616
617
618
619

620
621
622
623
624
625
626
611
612
613
614
615
616
617

618
619
620
621
622
623
624
625







-
+







			    (real-dir      (if (file-exists? run-dir)
					       (resolve-pathname run-dir)
					       #f))
			    (test-state    (db:test-get-state new-test-dat))
			    (test-fulln    (db:test-get-fullname new-test-dat)))
			   (case action
			     ((remove-runs)
			      (debug:print-info 0 "test: " test-name " item path: " item-path " test-state: " test-state)
			      (debug:print-info 0 "test: " test-name " itest-state: " test-state)
			      (if (member test-state (list "RUNNING" "LAUNCHED" "REMOTEHOSTSTART" "KILLREQ"))
				  (begin
				    (if (not (hash-table-ref/default test-retry-time test-fulln #f))
					(hash-table-set! test-retry-time test-fulln (current-seconds)))
				    (if (> (- (current-seconds)(hash-table-ref test-retry-time test-fulln)) allow-run-time)
				      ;; This test is not in a correct state for cleaning up. Let's try some graceful shutdown steps first
				      ;; Set the test to "KILLREQ" and wait five seconds then try again. Repeat up to five times then give