39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
|
39
40
41
42
43
44
45
46
47
48
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
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
117
118
119
120
121
122
|
+
+
-
-
-
-
-
-
+
+
+
+
+
+
-
-
-
-
-
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
|
;; Not currently used! But, I think it *should* be used!!!
(define (client:logout serverdat)
(let ((ok (and (socket? serverdat)
(cdb:logout serverdat *toppath* (client:get-signature)))))
ok))
;; BB: commenting out orphan code
;;;;;
(define (client:connect iface port)
(case (server:get-transport)
((rpc) (rpc:client-connect iface port))
((http) (http:client-connect iface port))
((zmq) (zmq:client-connect iface port))
(else (rpc:client-connect iface port))))
;; (define (client:connect iface port)
;; (case (server:get-transport)
;; ((rpc) (rpc:client-connect iface port))
;; ((http) (http:client-connect iface port))
;; ((zmq) (zmq:client-connect iface port))
;; (else (rpc:client-connect iface port))))
(define (client:setup run-id #!key (remaining-tries 10) (failed-connects 0))
(case (server:get-transport)
((rpc) (rpc-transport:client-setup run-id)) ;;(client:setup-rpc run-id))
((http)(client:setup-http run-id))
(else (rpc-transport:client-setup run-id)))) ;; (client:setup-rpc run-id))))
(define (client:setup run-id #!key (remaining-tries 10))
;;(BB> "Entered client:setup with run-id="run-id" and remaining-tries="remaining-tries)
(debug:print-info 2 *default-log-port* "client:setup remaining-tries=" remaining-tries)
(let* ((server-dat (tasks:bb-get-server-info run-id))
(transport (if (and server-dat (vector? server-dat)) (string->symbol (tasks:hostinfo-get-transport server-dat)) 'noserver)))
(case transport
((noserver) ;; no server registered
(BB> "noserver")
(if (<= remaining-tries 0)
(begin
(debug:print-error 0 *default-log-port* "failed to start or connect to server for run-id " run-id)
(exit 1))
(begin
(let ((num-available (tasks:bb-num-in-available-state run-id)))
(debug:print-info 0 *default-log-port* "client:setup, no server registered, remaining-tries=" remaining-tries " num-available=" num-available)
(if (< num-available 2)
(server:try-running run-id))
(thread-sleep! (+ 5 (random (- 20 remaining-tries)))) ;; give server a little time to start up, randomize a little to avoid start storms.
(client:setup run-id remaining-tries: (- remaining-tries 1))))))
((http) (BB> "have http") (client:setup-http run-id server-dat remaining-tries))
((rpc) (BB> "have rpc") (rpc-transport:client-setup run-id server-dat remtries: remaining-tries))
(else
(debug:print-error 0 *default-log-port* "(6) Transport ["
transport "] specified for run-id [" run-id "] is not implemented in client:setup. Cannot proceed.")
(exit 1)))))
;; client:setup-http
;;
;; For http transport, robustly ensure an advertised-running server is actually working and responding, and
;; establish tcp connection to server. For servers marked running but not responding, kill them and clear from mdb
;;
(define (client:setup-http run-id server-dat remaining-tries)
(let* ((iface (tasks:hostinfo-get-interface server-dat))
(hostname (tasks:hostinfo-get-hostname server-dat))
(port (tasks:hostinfo-get-port server-dat))
(start-res (http-transport:client-connect iface port))
(ping-res (rmt:login-no-auto-client-setup start-res run-id)))
(if (and start-res ping-res)
(begin
(rmt:set-cinfo run-id start-res) ;; (hash-table-set! *runremote* run-id start-res) ;; side-effect - *runremote* cache init fpr rmt:*
(debug:print-info 2 *default-log-port* "connected to " (http-transport:server-dat-make-url start-res))
start-res)
(begin ;; login failed but have a server record, clean out the record and try again
(debug:print-info 0 *default-log-port* "client:setup-http, login failed, will attempt to start server ... start-res=" start-res ", run-id=" run-id ", server-dat=" server-dat)
(http-transport:close-connections run-id)
(rmt:del-cinfo run-id) ;; (hash-table-delete! *runremote* run-id) ;; BB: suspect there is nothing to delete ...
(tasks:kill-server-run-id run-id) ;; -9 so the hung processes dont eat 100% when not responding to sigterm.
(tasks:bb-server-force-clean-run-record run-id iface port
" client:setup-http (server-dat = #t)")
(if (> remaining-tries 8)
(thread-sleep! (+ 1 (random 5))) ;; spread out the starts a little
(thread-sleep! (+ 15 (random 20)))) ;; it isn't going well. give it plenty of time
(server:try-running run-id)
(thread-sleep! 5) ;; give server a little time to start up
(client:setup run-id remaining-tries: (- remaining-tries 1))
))))
;; (define (client:login-no-auto-setup server-info run-id)
;; (case (server:get-transport)
;; ((rpc) (rpc:login-no-auto-client-setup server-info run-id))
;; ((http) (rmt:login-no-auto-client-setup server-info run-id))
;; (else (rpc:login-no-auto-client-setup server-info run-id))))
;;
|
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
|
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
|
-
+
|
;; (thread-sleep! 5)
;; (client:setup run-id remaining-tries: 10)) ;; (- remaining-tries 1)))
;; (begin
;; (debug:print 25 *default-log-port* "INFO: client:setup failed to connect, start-res=" start-res ", run-id=" run-id ", host-info=" host-info)
;; (thread-sleep! 5)
;; (client:setup run-id remaining-tries: (- remaining-tries 1))))))
;; ;; YUK: rename server-dat here
;; (let* ((server-dat (open-run-close tasks:get-server tasks:open-db run-id)))
;; (let* ((server-dat (open-run-close tasks:get-server-info tasks:open-db run-id)))
;; (debug:print-info 0 *default-log-port* "client:setup server-dat=" server-dat ", remaining-tries=" remaining-tries)
;; (if server-dat
;; (let* ((iface (tasks:hostinfo-get-interface server-dat))
;; (port (tasks:hostinfo-get-port server-dat))
;; (start-res (http-transport:client-connect iface port))
;; ;; (ping-res (server:ping-server run-id iface port))
;; (ping-res (rmt:login-no-auto-client-setup start-res run-id)))
|
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
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
|
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
|
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
+
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
+
+
+
+
+
|
;; 2. We are a run tests, list runs or other interactive process and we must figure out
;; *transport-type* and *runremote* from the monitor.db
;;
;; client:setup
;;
;; lookup_server, need to remove *runremote* stuff
;;
(define (client:setup-http run-id #!key (remaining-tries 10) (failed-connects 0))
(debug:print-info 2 *default-log-port* "client:setup remaining-tries=" remaining-tries)
(let* ((tdbdat (tasks:open-db)))
(if (<= remaining-tries 0)
(begin
(debug:print-error 0 *default-log-port* "failed to start or connect to server for run-id " run-id)
(exit 1))
(let* ((server-dat (tasks:get-server (db:delay-if-busy tdbdat) run-id)))
(debug:print-info 4 *default-log-port* "client:setup server-dat=" server-dat ", remaining-tries=" remaining-tries)
(if server-dat
(let* ((iface (tasks:hostinfo-get-interface server-dat))
(hostname (tasks:hostinfo-get-hostname server-dat))
(port (tasks:hostinfo-get-port server-dat))
(start-res (case *transport-type*
((http)(http-transport:client-connect iface port))
;;((nmsg)(nmsg-transport:client-connect hostname port))
))
(ping-res (case *transport-type*
((http)(rmt:login-no-auto-client-setup start-res run-id))
;; ((nmsg)(let ((logininfo (rmt:login-no-auto-client-setup start-res run-id)))
;; (if logininfo
;; (car (vector-ref logininfo 1))
;; #f)))
)))
(if (and start-res
ping-res)
(begin
(hash-table-set! *runremote* run-id start-res)
(debug:print-info 2 *default-log-port* "connected to " (http-transport:server-dat-make-url start-res))
start-res)
(begin ;; login failed but have a server record, clean out the record and try again
(debug:print-info 0 *default-log-port* "client:setup, login failed, will attempt to start server ... start-res=" start-res ", run-id=" run-id ", server-dat=" server-dat)
(case *transport-type*
((http)(http-transport:close-connections run-id)))
(hash-table-delete! *runremote* run-id)
(tasks:kill-server-run-id run-id)
(tasks:server-force-clean-run-record (db:delay-if-busy tdbdat)
run-id
(tasks:hostinfo-get-interface server-dat)
(tasks:hostinfo-get-port server-dat)
" client:setup (server-dat = #t)")
(if (> remaining-tries 8)
(thread-sleep! (+ 1 (random 5))) ;; spread out the starts a little
(thread-sleep! (+ 15 (random 20)))) ;; it isn't going well. give it plenty of time
(server:try-running run-id)
(thread-sleep! 5) ;; give server a little time to start up
(client:setup run-id remaining-tries: (- remaining-tries 1))
)))
(begin ;; no server registered
(let ((num-available (tasks:num-in-available-state (db:dbdat-get-db tdbdat) run-id)))
(debug:print-info 0 *default-log-port* "client:setup, no server registered, remaining-tries=" remaining-tries " num-available=" num-available)
(if (< num-available 2)
(server:try-running run-id))
(thread-sleep! (+ 5 (random (- 20 remaining-tries)))) ;; give server a little time to start up, randomize a little to avoid start storms.
(client:setup run-id remaining-tries: (- remaining-tries 1)))))))))
;; keep this as a function to ease future
(define (client:start run-id server-info)
(http-transport:client-connect (tasks:hostinfo-get-interface server-info)
(tasks:hostinfo-get-port server-info)))
|