Megatest

Check-in [0f348aa84b]
Login
Overview
Comment:Speculative fix for wrong server access issue
Downloads: Tarball | ZIP archive | SQL archive
Timelines: family | ancestors | descendants | both | v1.80
Files: files | file ages | folders
SHA1: 0f348aa84ba7889265c616f23b700efde5857538
User & Date: mrwellan on 2023-03-03 02:16:30
Other Links: branch diff | manifest | tags
Context
2023-03-03
17:37
wip check-in: abb2c73974 user: mrwellan tags: v1.80
02:16
Speculative fix for wrong server access issue check-in: 0f348aa84b user: mrwellan tags: v1.80
2023-03-02
21:32
Improve some of the logic around servers check-in: 87813c75d2 user: mrwellan tags: v1.80
Changes

Modified api.scm from [edf4eba965] to [1736f4bcb3].

246
247
248
249
250
251
252


253

254
255
256
257
258
259
260
261
262
263
264
265
266
267
268


269

270
271
272
273
274
275
276
246
247
248
249
250
251
252
253
254

255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272

273
274
275
276
277
278
279
280







+
+
-
+















+
+
-
+







	   (delay-wait (if (> newcount 10)
			   (- newcount 10)
			   0)))
      (set! *api-process-request-count* newcount)
      (set! *db-last-access* (current-seconds))
      (match indat
	((cmd run-id params meta)
	 (let* ((ttdat   *server-info*)
		(server-state (tt-state ttdat))
	 (let* ((status  (cond
		(status  (cond
			  ((> newcount 30) 'busy)
			  ((> newcount 15) 'loaded)
			  (else 'ok)))
		(errmsg  (case status
			   ((busy)   (conc "Server overloaded, "newcount" threads in flight"))
			   ((loaded) (conc "Server loaded, "newcount" threads in flight"))
			   (else     #f)))
		(result  (case status
			   ((busy)  (- newcount 29))
			   ((loaded) #f)
			   (else
			    (case cmd
			      ((ping) *server-signature*)
			      (else
			       (api:dispatch-request dbstruct cmd run-id params))))))
		(meta   (case cmd
			  ((ping) `((sstate . ,server-state)))
		(meta   `((wait . ,delay-wait)))
			  (else   `((wait . ,delay-wait)))))
		(payload (list status errmsg result meta)))
	   (set! *api-process-request-count* (- *api-process-request-count* 1))
	   (serialize payload)))
	(else
	 (assert #f "FATAL: failed to deserialize indat "indat))))))
       

Modified tcp-transportmod.scm from [6b42c5c5cc] to [2182e473e2].

82
83
84
85
86
87
88

89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105



106
107
108
109
110
111
112
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116







+

















+
+
+








;; Used for BOTH clients and servers
(defstruct tt
  ;; client related
  (conns (make-hash-table)) ;; dbfname -> conn

  ;; server related
  (state        'starting)
  (areapath     #f)
  (host         #f)
  (port         #f)
  (conn         #f)
  (cleanup-proc #f)
  (handler      #f) ;; receives data and responds
  (socket       #f)
  (thread       #f)
  (host-port    #f)
  (cmd-thread   #f)
  (ro-mode      #f)
  (ro-mode-checked #f)
  (last-access  (current-seconds))
  (servinf-file #f)
  (last-serv-start 0)
  )

;; make ttdat visible
(define *server-info* #f)

(define (tt:make-remote areapath)
  (make-tt areapath: areapath))

;; 1 ... or #f
(define (tt:valid-run-id run-id)
  (or (number? run-id)
      (not run-id)))
137
138
139
140
141
142
143
144
145
146
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
178
179
180
141
142
143
144
145
146
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
178
179
180

181
182

183
184
185
186
187
188
189
190







-
-
-
-
-
-
-
-
-
-
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+

















-
+

-
+







			   dbfname: dbfname
			   servinf-file: servinffile
			   server-id: server-id
			   server-start: start-time
			   pid: pid)))
	       (hash-table-set! (tt-conns ttdat) dbfname conn)
	       ;; verify we can talk to this server
	       (if (tt:ping host port server-id)
		   conn
		   (let* ((curr-secs (current-seconds)))
		     ;; rm the (last server) would go here
		     (if (> (- curr-secs (tt-last-serv-start ttdat)) 10)
			 (begin
			   (tt-last-serv-start-set! ttdat curr-secs)
			   (server-start-proc))) ;; don't try and start server unless 30 sec has gone by since last attempt
		     (thread-sleep! 1)
		     (tt:client-connect-to-server ttdat dbfname run-id testsuite)))))
	       (let* ((ping-res (tt:ping host port server-id)))
		 (case ping-res
		   ((running) conn)
		   ((starting)
		    (thread-sleep! 0.5)
		    (tt:client-connect-to-server ttdat dbfname run-id testsuite))
		   (else
		    (let* ((curr-secs (current-seconds)))
		      ;; rm the (last server) would go here
		      (if (> (- curr-secs (tt-last-serv-start ttdat)) 10)
			  (begin
			    (tt-last-serv-start-set! ttdat curr-secs)
			    (server-start-proc))) ;; start server if 30 sec since last attempt
		      (thread-sleep! 1)
		      (tt:client-connect-to-server ttdat dbfname run-id testsuite)))))))
		   
	    (else
	     (if (> (- (current-seconds) (tt-last-serv-start ttdat)) 5) ;; really do not want to swamp the machine with servers
		 (begin
		   (debug:print-info 0 *default-log-port* "No server found. Starting one for run-id "run-id" in dbfile "dbfname)
		   (server-start-proc)
		   (tt-last-serv-start-set! ttdat (current-seconds))))
	     (thread-sleep! 1)
	     (tt:client-connect-to-server ttdat dbfname run-id testsuite)))))))
    
(define (tt:ping host port server-id)
  (let*  ((res (tt:send-receive-direct host port `(ping #f #f #f)))) ;; please send me your server-id
    ;;
    ;; need two threads, one a 5 second timer
    ;;
    (match res
      ((status errmsg result meta)
       (if (equal? result server-id)
	   (begin
	   (let* ((server-state (alist-ref 'sstate meta)))
	     ;; (debug:print 0 *default-log-port* "Ping to "host":"port" successful.")
	     #t) ;; then we are good
	     (or server-state 'unk)) ;; then we are good
	   (begin
	     (debug:print 0 *default-log-port* "WARNING: server-id does not match, expected: "server-id", got: "result)
	     #f)))
      (else
       ;; (debug:print 0 *default-log-port* "res not in form (status errmsg result meta), got: "res)
       #f))))

351
352
353
354
355
356
357

358
359
360
361
362
363
364
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375







+







	 (nosyncdbpath (conc areapath"/.megatest"))
	 (cleanup (lambda ()
		    (if (tt-cleanup-proc ttdat)
			((tt-cleanup-proc ttdat)))
		    (dbfile:with-no-sync-db nosyncdbpath
					    (lambda (db)
					      (db:no-sync-del! db dbfname))))))
    (set! *server-info* ttdat)
    (let loop ((count 0))
      (if (> count 240)
	  (begin
	    (debug:print 0 *default-log-port* "FATAL: Could not start a tcp server, giving up.")
	    (exit 1))
	  (if (not (tt-port ttdat)) ;; no connection yet
	      (begin
383
384
385
386
387
388
389

390
391
392
393
394
395
396
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408







+







			    #t
			    (let* ((lockinfo  (dbfile:with-no-sync-db nosyncdbpath
								      (lambda (db)
									(db:no-sync-get-lock db dbfname))))
				   (success   (car lockinfo)))
			      (if success
				  (begin
				    (tt-state-set! ttdat 'running)
				    (debug:print 0 *default-log-port* "Got server lock for "dbfname)
				    (set! db-locked-in #t)
				    #t)
				  (begin
				    (debug:print 0 *default-log-port* "Failed to get server lock for "dbfname)
				    #f)))))
		       (else
468
469
470
471
472
473
474

475
476
477
478
479
480
481
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494







+







;;   (thread-join! (udat-cmd-thread uconn))
;;   (tcp-close (udat-socket uconn)))
;; 
;; 

(define (tt:shutdown-server ttdat)
  (let* ((cleanproc (tt-cleanup-proc ttdat)))
    (tt-state-set! ttdat 'shutdown)
    (if cleanproc (cleanproc))
    (tcp-close (tt-socket ttdat)) ;; close up ports here
    ))

;; (define (wait-and-close uconn)
;;   (thread-join! (tt-cmd-thread uconn))
;;   (tcp-close (tt-socket uconn)))