| ︙ | | |
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
|
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
|
-
+
|
(define *nmsg-mutex* (make-mutex))
;;======================================================================
;; S E R V E R
;;======================================================================
(define (nmsg-transport:run dbstruct hostn run-id server-id #!key (retrynum 1000))
(debug:print 2 "Attempting to start the server ...")
(debug:print 2 #f "Attempting to start the server ...")
(let* ((start-port (portlogger:open-run-close portlogger:find-port))
(server-thread (make-thread (lambda ()
(nmsg-transport:try-start-server dbstruct run-id start-port server-id))
"server thread"))
(tdbdat (tasks:open-db)))
(thread-start! server-thread)
(thread-sleep! 0.1)
|
| ︙ | | |
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
|
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
|
-
+
-
+
-
+
-
+
|
(tasks:server-set-state! (db:delay-if-busy tdbdat) server-id "running")
(thread-start! (make-thread
(lambda ()(nmsg-transport:keep-running server-id run-id))
"keep running"))
(thread-join! server-thread))
(if (> retrynum 0)
(begin
(debug:print 0 "WARNING: Failed to connect to server (self) on host " hostn ":" start-port ", trying again.")
(debug:print 0 #f "WARNING: Failed to connect to server (self) on host " hostn ":" start-port ", trying again.")
(tasks:server-delete-record (db:delay-if-busy tdbdat) server-id "failed to start, never received server alive signature")
(portlogger:open-run-close portlogger:set-failed start-port)
(nmsg-transport:run dbstruct hostn run-id server-id))
(begin
(debug:print 0 "ERROR: could not find an open port to start server on. Giving up")
(debug:print 0 #f "ERROR: could not find an open port to start server on. Giving up")
(exit 1))))))
(define (nmsg-transport:try-start-server dbstruct run-id portnum server-id)
(let ((repsoc (nn-socket 'rep)))
(nn-bind repsoc (conc "tcp://*:" portnum))
(let loop ((msg-in (nn-recv repsoc)))
(let* ((dat (db:string->obj msg-in transport: 'nmsg)))
(debug:print 0 "server, received: " dat)
(debug:print 0 #f "server, received: " dat)
(let ((result (api:execute-requests dbstruct dat)))
(debug:print 0 "server, sending: " result)
(debug:print 0 #f "server, sending: " result)
(nn-send repsoc (db:obj->string result transport: 'nmsg)))
(loop (nn-recv repsoc))))))
;; all routes though here end in exit ...
;;
(define (nmsg-transport:launch run-id)
(let* ((tdbdat (tasks:open-db))
|
| ︙ | | |
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
|
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
|
-
+
-
+
-
+
|
;; (daemon:ize)
;; (if *alt-log-file* ;; we should re-connect to this port, I think daemon:ize disrupts it
;; (begin
;; (current-error-port *alt-log-file*)
;; (current-output-port *alt-log-file*)))))
(if (server:check-if-running run-id)
(begin
(debug:print-info 0 "Server for run-id " run-id " already running")
(debug:print-info 0 #f "Server for run-id " run-id " already running")
(exit 0)))
(let loop ((server-id (tasks:server-lock-slot (db:delay-if-busy tdbdat) run-id))
(remtries 4))
(if (not server-id)
(if (> remtries 0)
(begin
(thread-sleep! 2)
(if (not (server:check-if-running run-id))
(loop (tasks:server-lock-slot (db:delay-if-busy tdbdat) run-id)
(- remtries 1))
(begin
(debug:print-info 0 "Another server took the slot, exiting")
(debug:print-info 0 #f "Another server took the slot, exiting")
(exit 0))))
(begin
;; since we didn't get the server lock we are going to clean up and bail out
(debug:print-info 2 "INFO: server pid=" (current-process-id) ", hostname=" (get-host-name) " not starting due to other candidates ahead in start queue")
(debug:print-info 2 #f "INFO: server pid=" (current-process-id) ", hostname=" (get-host-name) " not starting due to other candidates ahead in start queue")
(tasks:server-delete-records-for-this-pid (db:delay-if-busy tdbdat) " http-transport:launch")
))
;; locked in a server id, try to start up
(nmsg-transport:run dbstruct hostn run-id server-id))
(set! *didsomething* #t)
(exit))))
|
| ︙ | | |
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
|
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
|
-
+
|
(dat (vector "ping" our-key))
(result (condition-case
(nmsg-transport:client-api-send-receive-raw req dat timeout: timeout)
((timeout)(set! success #f) #f)))
(key (if success
(vector-ref result 1)
#f)))
(debug:print 0 "success=" success ", key=" key ", expected-key=" expected-key ", equal? " (equal? key expected-key))
(debug:print 0 #f "success=" success ", key=" key ", expected-key=" expected-key ", equal? " (equal? key expected-key))
(if (and success
(or (not expected-key) ;; just getting a reply is good enough then
(equal? key expected-key)))
(if return-socket
req
(begin
(if (not socket)(nn-close req)) ;; don't want a side effect of closing socket if handed it
|
| ︙ | | |
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
|
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
|
-
+
|
(set! success #t)
(set! result (db:string->obj res transport: 'nmsg))))
"send-recv"))
(timeout (make-thread
(lambda ()
(let loop ((count 0))
(thread-sleep! 1)
(debug:print-info 1 "send-receive-raw, still waiting after " count " seconds...")
(debug:print-info 1 #f "send-receive-raw, still waiting after " count " seconds...")
(if (and keepwaiting (< count timeout)) ;; yes, this is very aproximate
(loop (+ count 1))))
(if keepwaiting
(begin
(print "timeout waiting for ping")
(thread-terminate! send-recv))))
"timeout")))
|
| ︙ | | |
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
|
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
|
-
-
+
+
-
+
-
+
|
(if success (thread-terminate! timeout)))
;; raise timeout error if timed out
(if success
(if (and (vector? result)
(vector-ref result 0)) ;; did it fail at the server?
result ;; nope, all good
(begin
(debug:print 0 "ERROR: error occured at server, info=" (vector-ref result 2))
(debug:print 0 " client call chain:")
(debug:print 0 #f "ERROR: error occured at server, info=" (vector-ref result 2))
(debug:print 0 #f " client call chain:")
(print-call-chain (current-error-port))
(debug:print 0 " server call chain:")
(debug:print 0 #f " server call chain:")
(pp (vector-ref result 1) (current-error-port))
(signal (vector-ref result 0))))
(signal (make-composite-condition
(make-property-condition 'timeout 'message "nmsg-transport:client-api-send-receive-raw timed out talking to server"))))))
;; run nmsg-transport:keep-running in a parallel thread to monitor that the db is being
;; used and to shutdown after sometime if it is not.
;;
(define (nmsg-transport:keep-running server-id run-id)
;; if none running or if > 20 seconds since
;; server last used then start shutdown
;; This thread waits for the server to come alive
(let* ((server-info (let loop ()
(let ((sdat #f))
(mutex-lock! *heartbeat-mutex*)
(set! sdat *server-info*)
(mutex-unlock! *heartbeat-mutex*)
(if sdat
(begin
(debug:print-info 0 "keep-running got sdat=" sdat)
(debug:print-info 0 #f "keep-running got sdat=" sdat)
sdat)
(begin
(thread-sleep! 0.5)
(loop))))))
(iface (car server-info))
(port (cadr server-info))
(last-access 0)
|
| ︙ | | |
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
|
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
|
-
+
-
+
-
+
|
(set! last-access (http-transport:server-dat-get-last-access server-info)) ;; *last-db-access*)
(mutex-unlock! *heartbeat-mutex*)
(db:sync-touched *inmemdb* run-id force-sync: #t)
(if (and *server-run*
(> (+ last-access server-timeout)
(current-seconds)))
(begin
(debug:print-info 0 "Server continuing, seconds since last db access: " (- (current-seconds) last-access))
(debug:print-info 0 #f "Server continuing, seconds since last db access: " (- (current-seconds) last-access))
(loop 0))
(begin
(debug:print-info 0 "Starting to shutdown the server.")
(debug:print-info 0 #f "Starting to shutdown the server.")
(set! *time-to-exit* #t)
(db:sync-touched *inmemdb* run-id force-sync: #t)
(tasks:server-delete-record (db:delay-if-busy tdbdat) server-id " http-transport:keep-running")
(debug:print-info 0 "Server shutdown complete. Exiting")
(debug:print-info 0 #f "Server shutdown complete. Exiting")
(exit)
))))))
;;======================================================================
;; C L I E N T S
;;======================================================================
|
| ︙ | | |
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
|
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
|
-
+
-
+
-
+
|
;;======================================================================
;; DO NOT USE
;;
(define (nmsg-transport:client-signal-handler signum)
(handle-exceptions
exn
(debug:print " ... exiting ...")
(debug:print 0 #f " ... exiting ...")
(let ((th1 (make-thread (lambda ()
(if (not *received-response*)
(receive-message* *runremote*))) ;; flush out last call if applicable
"eat response"))
(th2 (make-thread (lambda ()
(debug:print 0 "ERROR: Received ^C, attempting clean exit. Please be patient and wait a few seconds before hitting ^C again.")
(debug:print 0 #f "ERROR: Received ^C, attempting clean exit. Please be patient and wait a few seconds before hitting ^C again.")
(thread-sleep! 3) ;; give the flush three seconds to do it's stuff
(debug:print 0 " Done.")
(debug:print 0 #f " Done.")
(exit 4))
"exit on ^C timer")))
(thread-start! th2)
(thread-start! th1)
(thread-join! th2))))
|