Overview
Comment: | Converted a few load wait back to simple sleeps to save cycles. |
---|---|
Downloads: | Tarball | ZIP archive | SQL archive |
Timelines: | family | ancestors | descendants | both | v1.65-cleanup | v1.6565 |
Files: | files | file ages | folders |
SHA1: |
44f224d1e87cfadc4de688c702dc942a |
User & Date: | mrwellan on 2020-08-28 10:24:31 |
Other Links: | branch diff | manifest | tags |
Context
2020-08-28
| ||
14:24 | lock with host/pid check-in: 9657acc77f user: matt tags: v1.65-cleanup | |
10:24 | Converted a few load wait back to simple sleeps to save cycles. check-in: 44f224d1e8 user: mrwellan tags: v1.65-cleanup, v1.6565 | |
2020-08-27
| ||
14:45 | made server throttle wait time configurable ==3.4/1.1/1203/PASS/mars== check-in: 2415e799b1 user: mrwellan tags: v1.65-cleanup, v1.6565 | |
Changes
Modified runs.scm from [3655db2bdd] to [dce077e9b3].
︙ | ︙ | |||
1263 1264 1265 1266 1267 1268 1269 | (if (or (null? fails) (member 'toplevel testmode)) (begin ;; couldn't run, take a breather (if (runs:lownoise "Waiting for more work to do..." 60) (debug:print-info 0 *default-log-port* "Waiting for more work to do...")) | | | | 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 | (if (or (null? fails) (member 'toplevel testmode)) (begin ;; couldn't run, take a breather (if (runs:lownoise "Waiting for more work to do..." 60) (debug:print-info 0 *default-log-port* "Waiting for more work to do...")) ;; (if (runs:dat-load-mgmt-function runsdat)((runs:dat-load-mgmt-function runsdat))) (thread-sleep! 5) (list (car newtal)(cdr newtal) reg reruns)) ;; the waiton is FAIL so no point in trying to run hed ever again (begin (let ((my-test-id (rmt:get-test-id run-id test-name item-path))) (mt:test-set-state-status-by-id-unless-completed run-id my-test-id "COMPLETED" "PREQ_FAIL" "Failed to run due to failed prerequisites2")) (if (or (not (null? reg))(not (null? tal))) (if (vector? hed) |
︙ | ︙ | |||
1798 1799 1800 1801 1802 1803 1804 | (debug:print-info 4 *default-log-port* "Exiting loop with...\n hed=" hed "\n tal=" tal "\n reruns=" reruns)) ))) ;; end loop on sorted test names ;; this is the point where everything is launched and now you can mark the run in metadata table as all launched (rmt:set-var (conc "lunch-complete-" run-id) "yes") ;; now *if* -run-wait we wait for all tests to be done ;; Now wait for any RUNNING tests to complete (if in run-wait mode) | | | | | 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 | (debug:print-info 4 *default-log-port* "Exiting loop with...\n hed=" hed "\n tal=" tal "\n reruns=" reruns)) ))) ;; end loop on sorted test names ;; this is the point where everything is launched and now you can mark the run in metadata table as all launched (rmt:set-var (conc "lunch-complete-" run-id) "yes") ;; now *if* -run-wait we wait for all tests to be done ;; Now wait for any RUNNING tests to complete (if in run-wait mode) ;; (if (runs:dat-load-mgmt-function runsdat)((runs:dat-load-mgmt-function runsdat))) (thread-sleep! 10) ;; I think there is a race condition here. Let states/statuses settle (let wait-loop ((num-running (rmt:get-count-tests-running-for-run-id run-id #t)) ;; fastmode=yes (prev-num-running 0)) ;; (debug:print-info 13 *default-log-port* "num-running=" num-running ", prev-num-running=" prev-num-running) (if (and (or (args:get-arg "-run-wait") (equal? (configf:lookup *configdat* "setup" "run-wait") "yes")) (> num-running 0)) (begin ;; Here we mark any old defunct tests as incomplete. Do this every fifteen minutes ;; (debug:print 0 *default-log-port* "Got here eh! num-running=" num-running " (> num-running 0) " (> num-running 0)) (if (> (current-seconds)(+ last-time-incomplete 900)) (let ((actual-num-running (rmt:get-count-tests-running-for-run-id run-id #f))) ;; fastmode=no (debug:print-info 0 *default-log-port* "Marking stuck tests as INCOMPLETE while waiting for run " run-id ". Running as pid " (current-process-id) " on " (get-host-name)) (set! last-time-incomplete (current-seconds)) ;; FIXME, this might be causing slow down - use of set! (rmt:find-and-mark-incomplete run-id #f) (debug:print-info 0 *default-log-port* "run-wait specified, waiting on " actual-num-running " tests in RUNNING, REMOTEHOSTSTART or LAUNCHED state at " (time->string (seconds->local-time (current-seconds)))))) ;; (if (runs:dat-load-mgmt-function runsdat)((runs:dat-load-mgmt-function runsdat))) (thread-sleep! 5) ;; (if (>= num-running max-concurrent-jobs) 5 1)) (wait-loop (rmt:get-count-tests-running-for-run-id run-id #t) ;; fastmode=yes num-running)))) ;; LET* ((test-record ;; we get here on "drop through". All done! ;; this is moved to runs:run-testes since this function is getting called twice to ensure everthing is completed. ;; (debug:print-info 0 *default-log-port* "Calling Post Hook") ;; (runs:run-post-hook run-id) |
︙ | ︙ | |||
1957 1958 1959 1960 1961 1962 1963 | (debug:print 2 *default-log-port* "WARN: Test not pre-created? test-name=" test-name ", item-path=" item-path ", run-id=" run-id) (rmt:register-test run-id test-name item-path) (set! test-id (rmt:get-test-id run-id test-name item-path)))) (debug:print-info 4 *default-log-port* "test-id=" test-id ", run-id=" run-id ", test-name=" test-name ", item-path=\"" item-path "\"") (set! testdat (rmt:get-test-info-by-id run-id test-id)) (if (not testdat) (begin | | | | | 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 | (debug:print 2 *default-log-port* "WARN: Test not pre-created? test-name=" test-name ", item-path=" item-path ", run-id=" run-id) (rmt:register-test run-id test-name item-path) (set! test-id (rmt:get-test-id run-id test-name item-path)))) (debug:print-info 4 *default-log-port* "test-id=" test-id ", run-id=" run-id ", test-name=" test-name ", item-path=\"" item-path "\"") (set! testdat (rmt:get-test-info-by-id run-id test-id)) (if (not testdat) (begin (debug:print-info 0 *default-log-port* "WARNING: server is overloaded, trying again in two seconds") ;; (if (runs:dat-load-mgmt-function runsdat)((runs:dat-load-mgmt-function runsdat))) (thread-sleep! 2) (loop))))) (if (not testdat) ;; should NOT happen (debug:print-error 0 *default-log-port* "failed to get test record for test-id " test-id)) (set! test-id (db:test-get-id testdat)) (if (common:file-exists? test-path) (change-directory test-path) (begin |
︙ | ︙ |