49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
|
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
|
-
-
+
+
-
-
-
-
-
-
-
-
-
+
+
+
+
+
+
+
+
+
+
+
|
(define (rmt:write-frequency-over-limit? cmd run-id)
(and (not (member cmd api:read-only-queries))
(let* ((tmprec (hash-table-ref/default *write-frequency* run-id #f))
(record (if tmprec tmprec
(let ((v (vector (current-seconds) 0)))
(hash-table-set! *write-frequency* run-id v)
v)))
(count (+ 1 (vector-ref record 1)))
(start (vector-ref record 0))
(count (+ 1 (safe-vector-ref record 1)))
(start (safe-vector-ref record 0))
(queries-per-second (/ (* count 1.0)
(max (- (current-seconds) start) 1))))
(vector-set! record 1 count)
(if (and (> count 10)
(> queries-per-second 10))
(begin
(debug:print-info 1 "db write rate too high, starting a server, count=" count " start=" start " run-id=" run-id " queries-per-second=" queries-per-second)
#t)
#f))))
(define (rmt:get-connection-info run-id)
(let ((cinfo (hash-table-ref/default *runremote* run-id #f)))
(if cinfo
cinfo
;; NB// can cache the answer for server running for 10 seconds ...
;; ;; (and (not (rmt:write-frequency-over-limit? cmd run-id))
(if (tasks:server-running-or-starting? (db:delay-if-busy (tasks:open-db)) run-id)
(client:setup run-id)
#f))))
;; (define (rmt:get-connection-info run-id)
;; (let ((cinfo (hash-table-ref/default *runremote* run-id #f)))
;; (if cinfo
;; cinfo
;; ;; NB// can cache the answer for server running for 10 seconds ...
;; ;; ;; (and (not (rmt:write-frequency-over-limit? cmd run-id))
;; ;; (if (tasks:server-running-or-starting? (db:delay-if-busy (tasks:open-db)) run-id)
;; ;; (begin
;; ;; (tasks:start-and-wait-for-server (db:delay-if-busy (tasks:open-db)) run-id)
;; (client:setup run-id))))
;; ;; #f))))
(define *send-receive-mutex* (make-mutex)) ;; should have separate mutex per run-id
(define (rmt:send-receive cmd rid params #!key (attemptnum 1)) ;; start attemptnum at 1 so the modulo below works as expected
;; clean out old connections
(mutex-lock! *db-multi-sync-mutex*)
(let ((expire-time (- (current-seconds) (server:get-timeout) 10))) ;; don't forget the 10 second margin
(for-each
|
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
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
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
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
|
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
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
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
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
|
-
+
+
+
-
-
+
+
+
+
-
-
-
+
+
+
+
-
+
+
-
-
-
-
-
-
-
-
-
-
-
-
-
+
-
-
-
-
-
-
+
+
-
+
-
-
+
+
-
-
+
+
|
((nmsg)(nn-close (http-transport:server-dat-get-socket
(hash-table-ref *runremote* run-id)))))
(hash-table-delete! *runremote* run-id)))))
(hash-table-keys *runremote*)))
(mutex-unlock! *db-multi-sync-mutex*)
;; (mutex-lock! *send-receive-mutex*)
(let* ((run-id (if rid rid 0))
(connection-info (rmt:get-connection-info run-id)))
(connection-info (hash-table-ref/default *runremote* run-id #f))) ;; (rmt:get-connection-info run-id)))
;; the nmsg method does the encoding under the hood (the http method should be changed to do this also)
(if connection-info
;; use the server if have connection info
(let* ((dat (case *transport-type*
((http)(condition-case
(http-transport:client-api-send-receive run-id connection-info cmd params)
((commfail)
(tasks:kill-server-run-id run-id)
((commfail)(vector #f "communications fail"))
((exn)(vector #f "other fail"))))
(vector #f "communications fail"))
((exn)
(tasks:kill-server-run-id run-id)
(vector #f "other fail"))))
((nmsg)(condition-case
(nmsg-transport:client-api-send-receive run-id connection-info cmd params)
((timeout)(vector #f "timeout talking to server"))))
(else (exit))))
(success (if (vector? dat) (vector-ref dat 0) #f))
(res (if (vector? dat) (vector-ref dat 1) #f)))
(if (vector? connection-info)(http-transport:server-dat-update-last-access connection-info))
(success (if (vector? dat) (safe-vector-ref dat 0) #f))
(res (if (vector? dat) (safe-vector-ref dat 1) #f)))
(if (and connection-info (vector? connection-info))(http-transport:server-dat-update-last-access connection-info))
(if success
(begin
;; (mutex-unlock! *send-receive-mutex*)
;; all is well, return the result!
(case *transport-type*
((http) res) ;; (db:string->obj res))
((nmsg) res))) ;; (vector-ref res 1)))
((nmsg) res))) ;; (safe-vector-ref res 1)))
;; we had a connection but it is borked. clean up and reconnect
(begin ;; let ((new-connection-info (client:setup run-id)))
(debug:print 0 "WARNING: Communication failed, trying call to rmt:send-receive again.")
;; (case *transport-type*
;; ((nmsg)(nn-close (http-transport:server-dat-get-socket connection-info))))
(hash-table-delete! *runremote* run-id) ;; don't keep using the same connection
;; NOTE: killing server causes this process to block forever. No idea why. Dec 2.
;; (if (eq? (modulo attemptnum 5) 0)
;; (tasks:kill-server-run-id run-id tag: "api-send-receive-failed"))
;; (mutex-unlock! *send-receive-mutex*) ;; close the mutex here to allow other threads access to communications
(tasks:start-and-wait-for-server (tasks:open-db) run-id 15)
;; (nmsg-transport:client-api-send-receive run-id connection-info cmd param remtries: (- remtries 1))))))
;; no longer killing the server in http-transport:client-api-send-receive
;; may kill it here but what are the criteria?
;; start with three calls then kill server
;; (if (eq? attemptnum 3)(tasks:kill-server-run-id run-id))
;; (thread-sleep! 2)
(rmt:send-receive cmd run-id params attemptnum: (+ attemptnum 1)))))
;; no connection info? try to start a server
(if (and (< attemptnum 15)
(member cmd api:write-queries))
(begin
(hash-table-delete! *runremote* run-id)
;; (mutex-unlock! *send-receive-mutex*)
(tasks:start-and-wait-for-server (db:delay-if-busy (tasks:open-db)) run-id 10)
;; (client:setup run-id) ;; client setup happens in rmt:get-connection-info
(client:setup run-id)
(thread-sleep! (random 5)) ;; give some time to settle and minimize collison?
(rmt:send-receive cmd rid params attemptnum: (+ attemptnum 1)))
(begin
;; (debug:print 0 "ERROR: Communication failed!")
;; (mutex-unlock! *send-receive-mutex*)
;; (exit)
(rmt:open-qry-close-locally cmd run-id params)
)))))
(define (rmt:update-db-stats run-id rawcmd params duration)
(mutex-lock! *db-stats-mutex*)
(handle-exceptions
exn
(begin
(debug:print 0 "WARNING: stats collection failed in update-db-stats")
(debug:print 0 " message: " ((condition-property-accessor 'exn 'message) exn))
(print "exn=" (condition->list exn))
#f) ;; if this fails we don't care, it is just stats
(let* ((cmd (conc "run-id=" run-id " " (if (eq? rawcmd 'general-call) (car params) rawcmd)))
(stat-vec (hash-table-ref/default *db-stats* cmd #f)))
(if (not (vector? stat-vec))
(let ((newvec (vector 0 0)))
(hash-table-set! *db-stats* cmd newvec)
(set! stat-vec newvec)))
(vector-set! stat-vec 0 (+ (vector-ref stat-vec 0) 1))
(vector-set! stat-vec 1 (+ (vector-ref stat-vec 1) duration))))
(vector-set! stat-vec 0 (+ (safe-vector-ref stat-vec 0) 1))
(vector-set! stat-vec 1 (+ (safe-vector-ref stat-vec 1) duration))))
(mutex-unlock! *db-stats-mutex*))
(define (rmt:print-db-stats)
(let ((fmtstr "~40a~7-d~9-d~20,2-f")) ;; "~20,2-f"
(debug:print 18 "DB Stats\n========")
(debug:print 18 (format #f "~40a~8a~10a~10a" "Cmd" "Count" "TotTime" "Avg"))
(for-each (lambda (cmd)
(let ((cmd-dat (hash-table-ref *db-stats* cmd)))
(debug:print 18 (format #f fmtstr cmd (vector-ref cmd-dat 0) (vector-ref cmd-dat 1) (/ (vector-ref cmd-dat 1)(vector-ref cmd-dat 0))))))
(debug:print 18 (format #f fmtstr cmd (safe-vector-ref cmd-dat 0) (safe-vector-ref cmd-dat 1) (/ (safe-vector-ref cmd-dat 1)(safe-vector-ref cmd-dat 0))))))
(sort (hash-table-keys *db-stats*)
(lambda (a b)
(> (vector-ref (hash-table-ref *db-stats* a) 0)
(vector-ref (hash-table-ref *db-stats* b) 0)))))))
(> (safe-vector-ref (hash-table-ref *db-stats* a) 0)
(safe-vector-ref (hash-table-ref *db-stats* b) 0)))))))
(define (rmt:get-max-query-average run-id)
(mutex-lock! *db-stats-mutex*)
(let* ((runkey (conc "run-id=" run-id " "))
(cmds (filter (lambda (x)
(substring-index runkey x))
(hash-table-keys *db-stats*)))
(res (if (null? cmds)
(cons 'none 0)
(let loop ((cmd (car cmds))
(tal (cdr cmds))
(max-cmd (car cmds))
(res 0))
(let* ((cmd-dat (hash-table-ref *db-stats* cmd))
(tot (vector-ref cmd-dat 0))
(curravg (/ (vector-ref cmd-dat 1) (vector-ref cmd-dat 0))) ;; count is never zero by construction
(tot (safe-vector-ref cmd-dat 0))
(curravg (/ (safe-vector-ref cmd-dat 1) (safe-vector-ref cmd-dat 0))) ;; count is never zero by construction
(currmax (max res curravg))
(newmax-cmd (if (> curravg res) cmd max-cmd)))
(if (null? tal)
(if (> tot 10)
(cons newmax-cmd currmax)
(cons 'none 0))
(loop (car tal)(cdr tal) newmax-cmd currmax)))))))
|
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
|
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
|
-
+
-
+
-
+
|
(db (make-dbr:dbstruct path: dbdir local: #t)))
(set! *dbstruct-db* db)
db)))
(db-file-path (db:dbfile-path 0)))
;; (read-only (not (file-read-access? db-file-path)))
(let* ((start (current-milliseconds))
(resdat (api:execute-requests dbstruct-local (vector (symbol->string cmd) params)))
(res (vector-ref resdat 1))
(res (safe-vector-ref resdat 1))
(duration (- (current-milliseconds) start)))
(rmt:update-db-stats run-id cmd params duration)
;; mark this run as dirty if this was a write
(if (not (member cmd api:read-only-queries))
(let ((start-time (current-seconds)))
(mutex-lock! *db-multi-sync-mutex*)
;; (if (not (hash-table-ref/default *db-local-sync* run-id #f))
;; just set it every time. Is a write more expensive than a read and does it matter?
(hash-table-set! *db-local-sync* (or run-id 0) start-time) ;; the oldest "write"
(mutex-unlock! *db-multi-sync-mutex*)))
res)))
(define (rmt:send-receive-no-auto-client-setup connection-info cmd run-id params)
(let* ((run-id (if run-id run-id 0))
;; (jparams (db:obj->string params)) ;; (rmt:dat->json-str params))
(res (handle-exceptions
exn
#f
(http-transport:client-api-send-receive run-id connection-info cmd params))))
;; ((commfail) (vector #f "communications fail")))))
(if (and res (vector-ref res 0))
(if (and res (safe-vector-ref res 0))
res
#f)))
;; (db:string->obj (vector-ref dat 1))
;; (db:string->obj (safe-vector-ref dat 1))
;; (begin
;; (debug:print 0 "ERROR: rmt:send-receive-no-auto-client-setup failed, attempting to continue. Got " dat)
;; dat))))
;; Wrap json library for strings (why the ports crap in the first place?)
(define (rmt:dat->json-str dat)
(with-output-to-string
|