Overview
| Comment: | Turn off waiting in kind-run, rely on gating from wait-for-server-start-last-flag |
|---|---|
| Downloads: | Tarball | ZIP archive | SQL archive |
| Timelines: | family | ancestors | descendants | both | v1.65 |
| Files: | files | file ages | folders |
| SHA1: |
889e2e71ef5e685aa8ce4250ae038d57 |
| User & Date: | matt on 2021-03-12 14:25:17 |
| Other Links: | branch diff | manifest | tags |
Context
|
2021-03-12
| ||
| 18:53 | Backed out cleanup of config: functions check-in: ae1ac4c3b4 user: mmgraham tags: v1.65 | |
| 14:25 | Turn off waiting in kind-run, rely on gating from wait-for-server-start-last-flag check-in: 889e2e71ef user: matt tags: v1.65 | |
|
2021-03-11
| ||
| 17:52 | updated delay to kind-run check-in: dd2cea12eb user: pjhatwal tags: v1.65, THIS-IS-REAL-1.65 | |
Changes
Modified server.scm from [33c77942b5] to [19fd15cc14].
| ︙ | ︙ | |||
200 201 202 203 204 205 206 |
(string->number (cadr dat)) ;; port
(string->number (caddr dat))
(cadr (cddr dat))))))
(begin
(if dbprep-found
(begin
(debug:print-info 0 *default-log-port* "Server is in dbprep at " (current-seconds))
| | | 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 |
(string->number (cadr dat)) ;; port
(string->number (caddr dat))
(cadr (cddr dat))))))
(begin
(if dbprep-found
(begin
(debug:print-info 0 *default-log-port* "Server is in dbprep at " (current-seconds))
(thread-sleep! 0.5) ;; was 25 sec but that blocked things from starting?
)
(debug:print-info 0 *default-log-port* "Unable to get server info from " logf " at " (current-seconds))
)
(list #f #f #f #f)))))))))
;; get a list of servers with all relevant data
;; ( mod-time host port start-time pid )
|
| ︙ | ︙ | |||
403 404 405 406 407 408 409 |
;; look for $MT_RUN_AREA_HOME/logs/server-start-last
;; and wait for it to be at least 3 seconds old
(server:wait-for-server-start-last-flag areapath)
(if (not (server:check-if-running areapath)) ;; why try if there is already a server running?
(let* ((last-run-dat (hash-table-ref/default *server-kind-run* areapath '(0 0))) ;; callnum, whenrun
(call-num (car last-run-dat))
(when-run (cadr last-run-dat))
| | | | | | | | | | | > | | 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 |
;; look for $MT_RUN_AREA_HOME/logs/server-start-last
;; and wait for it to be at least 3 seconds old
(server:wait-for-server-start-last-flag areapath)
(if (not (server:check-if-running areapath)) ;; why try if there is already a server running?
(let* ((last-run-dat (hash-table-ref/default *server-kind-run* areapath '(0 0))) ;; callnum, whenrun
(call-num (car last-run-dat))
(when-run (cadr last-run-dat))
(run-delay (+ (case call-num ;; NOT USED. Waiting is handled by wait-for-server
((0) 0)
((1) 20)
((2) 300)
(else 600))
(random 5)
0)) ;; add a small random number just in case a lot of jobs hit the work hosts simultaneously
(lock-file (conc areapath "/logs/server-start.lock")))
;; (if (> (- (current-seconds) when-run) run-delay)
(let* ((start-flag (conc areapath "/logs/server-start-last")))
(common:simple-file-lock-and-wait lock-file expire-time: 15)
(debug:print-info 0 *default-log-port* "server:kind-run: touching " start-flag)
(system (conc "touch " start-flag)) ;; lazy but safe
(server:run areapath)
(thread-sleep! 2) ;; don't release the lock for at least a few seconds
(common:simple-file-release-lock lock-file)))
(debug:print-info 0 *default-log-port* "Found server already running. NOT trying to start another.")))
;; (hash-table-set! *server-kind-run* areapath (list (+ call-num 1)(current-seconds))))))
;; this one seems to be the general entry point
;;
(define (server:start-and-wait areapath #!key (timeout 60))
(let ((give-up-time (+ (current-seconds) timeout)))
(let loop ((server-info (server:check-if-running areapath))
(try-num 0))
|
| ︙ | ︙ |