1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
|
;; end_time,strftime('%s','now') as now from tests where state in
;; ('RUNNING','REMOTEHOSTSTART','LAUNCHED'));
(define (db:find-and-mark-incomplete dbstruct run-id ovr-deadtime)
(let* ((incompleted '())
(oldlaunched '())
(toplevels '())
(deadtime-str (configf:lookup *configdat* "setup" "deadtime")) ;; FIXME suspect test run time & deadtime are not well matched; resulting in COMPLETED/DEAD status of an a-ok running test
(deadtime (if (and deadtime-str
(string->number deadtime-str))
(string->number deadtime-str)
120))) ;; two minutes (30 seconds between updates, this leaves 3x grace period)
(if (number? ovr-deadtime)(set! deadtime ovr-deadtime))
"SELECT id,rundir,uname,testname,item_path FROM tests WHERE run_id=? AND (strftime('%s','now') - event_time) > (run_duration + ?) AND state IN ('RUNNING','REMOTEHOSTSTART');"
run-id deadtime)
;; in LAUNCHED for more than one day. Could be long due to job queues TODO/BUG: Need override for this in config
;;
;; (db:delay-if-busy dbdat)
(sqlite3:for-each-row
(lambda (test-id run-dir uname testname item-path)
(if (and (equal? uname "n/a")
(equal? item-path "")) ;; this is a toplevel test
;; what to do with toplevel? call rollup?
(set! toplevels (cons (list test-id run-dir uname testname item-path run-id) toplevels))
(set! oldlaunched (cons (list test-id run-dir uname testname item-path run-id) oldlaunched))))
db
"SELECT id,rundir,uname,testname,item_path FROM tests WHERE run_id=? AND (strftime('%s','now') - event_time) > 86400 AND state IN ('LAUNCHED');"
run-id)
(debug:print-info 18 *default-log-port* "Found " (length oldlaunched) " old LAUNCHED items, " (length toplevels) " old LAUNCHED toplevel tests and " (length incompleted) " tests marked RUNNING but apparently dead.")
;; These are defunct tests, do not do all the overhead of set-state-status. Force them to INCOMPLETE.
|
|
>
>
>
>
>
|
|
|
|
>
<
<
>
>
>
>
>
>
>
>
>
>
>
>
>
>
|
|
>
>
|
|
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
|
;; end_time,strftime('%s','now') as now from tests where state in
;; ('RUNNING','REMOTEHOSTSTART','LAUNCHED'));
(define (db:find-and-mark-incomplete dbstruct run-id ovr-deadtime)
(let* ((incompleted '())
(oldlaunched '())
(toplevels '())
(deadtime-trim (or ovr-deadtime (configf:lookup-number *configdat* "setup" "deadtime")))
(server-start-allowance 200)
(server-overloaded-budget 200)
(launch-monitor-off-time (or (configf:lookup-number *configdat* "setup" "test-stats-update-period") 30))
(launch-monitor-on-time-budget 30)
(launch-monitor-period (+ launch-monitor-off-time launch-monitor-on-time-budget server-overloaded-budget))
(remotehoststart-deadtime-default (+ server-start-allowance server-overloaded-budget 30))
(remotehoststart-deadtime (or deadtime-trim remotehoststart-deadtime-default))
(running-deadtime-default (+ server-start-allowance (* 2 launch-monitor-period)))
(running-deadtime (or deadtime-trim running-deadtime-default)) ;; two minutes (30 seconds between updates, this leaves 3x grace period)
)
(db:with-db
dbstruct #f #f
(lambda (db)
;; in RUNNING or REMOTEHOSTSTART for more than 10 minutes
;;
;; HOWEVER: this code in run:test seems to work fine
;; (> (- (current-seconds)(+ (db:test-get-event_time testdat)
;; (db:test-get-run_duration testdat)))
;; 600)
;; (db:delay-if-busy dbdat)
(sqlite3:for-each-row
(lambda (test-id run-dir uname testname item-path)
(if (and (equal? uname "n/a")
(equal? item-path "")) ;; this is a toplevel test
;; what to do with toplevel? call rollup?
(begin
(set! toplevels (cons (list test-id run-dir uname testname item-path run-id) toplevels))
(debug:print-info 0 *default-log-port* "Found old toplevel test in RUNNING state, test-id=" test-id))
(set! incompleted (cons (list test-id run-dir uname testname item-path run-id) incompleted))))
db
"SELECT id,rundir,uname,testname,item_path,event_time,run_duration FROM tests WHERE run_id=? AND (strftime('%s','now') - event_time) > (run_duration + ?) AND state IN ('RUNNING');"
run-id running-deadtime)
(sqlite3:for-each-row
(lambda (test-id run-dir uname testname item-path event-time run-duration)
(if (and (equal? uname "n/a")
(equal? item-path "")) ;; this is a toplevel test
;; what to do with toplevel? call rollup?
(begin
(set! toplevels (cons (list test-id run-dir uname testname item-path run-id) toplevels))
(debug:print-info 0 *default-log-port* "Found old toplevel test in RUNNING state, test-id=" test-id))
(begin
(debug:print-info 0 *default-log-port* "Found old test in RUNNING state, test-id=" test-id" exceeded running-deadtime "running-deadtime" now="(current-seconds)" event-time="event-time" run-duration="run-duration)
(set! incompleted (cons (list test-id run-dir uname testname item-path run-id) incompleted)))))
db
"SELECT id,rundir,uname,testname,item_path FROM tests WHERE run_id=? AND (strftime('%s','now') - event_time) > (run_duration + ?) AND state IN ('REMOTEHOSTSTART');"
run-id remotehoststart-deadtime)
;; in LAUNCHED for more than one day. Could be long due to job queues TODO/BUG: Need override for this in config
;;
;; (db:delay-if-busy dbdat)
(sqlite3:for-each-row
(lambda (test-id run-dir uname testname item-path)
(if (and (equal? uname "n/a")
(equal? item-path "")) ;; this is a toplevel test
;; what to do with toplevel? call rollup?
(set! toplevels (cons (list test-id run-dir uname testname item-path run-id) toplevels))
(begin
(debug:print-info 0 *default-log-port* "Found old test in REMOTEHOSTSTART state, test-id=" test-id" exceeded remotehoststart-deadtime "remotehoststart-deadtime)
(set! oldlaunched (cons (list test-id run-dir uname testname item-path run-id) oldlaunched)))))
db
"SELECT id,rundir,uname,testname,item_path FROM tests WHERE run_id=? AND (strftime('%s','now') - event_time) > 86400 AND state IN ('LAUNCHED');"
run-id)
(debug:print-info 18 *default-log-port* "Found " (length oldlaunched) " old LAUNCHED items, " (length toplevels) " old LAUNCHED toplevel tests and " (length incompleted) " tests marked RUNNING but apparently dead.")
;; These are defunct tests, do not do all the overhead of set-state-status. Force them to INCOMPLETE.
|