33
34
35
36
37
38
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
|
33
34
35
36
37
38
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
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
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
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
|
-
-
-
-
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
-
-
+
+
+
+
-
-
-
-
+
+
+
-
-
-
-
-
-
-
-
-
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
-
-
-
-
-
-
-
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
-
-
-
-
-
-
+
+
+
+
-
+
-
-
-
-
-
+
+
+
+
+
|
;; client:get-signature
(define (client:get-signature)
(if *my-client-signature* *my-client-signature*
(let ((sig (conc (get-host-name) " " (current-process-id))))
(set! *my-client-signature* sig)
*my-client-signature*)))
;; client:login serverdat
(define (client:login serverdat)
(cdb:login serverdat *toppath* (client:get-signature)))
;; 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))
(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: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))))
(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-rpc run-id)
;; (debug:print 0 "INFO: client:setup remaining-tries=" remaining-tries)
;; (if (<= remaining-tries 0)
;; (begin
;; (debug:print 0 "ERROR: failed to start or connect to server for run-id " run-id)
;; (exit 1))
;; (let ((host-info (hash-table-ref/default *runremote* run-id #f)))
;; (debug:print-info 0 "client:setup host-info=" host-info ", remaining-tries=" remaining-tries)
;; (if host-info
;; (let* ((iface (car host-info))
;; (port (cadr host-info))
;; (start-res (client:connect iface port))
;; ;; (ping-res (server:ping-server run-id iface port))
;; (ping-res (client:login-no-auto-setup start-res run-id)))
;; (if ping-res ;; sucessful login?
;; (begin
;; (hash-table-set! *runremote* run-id start-res)
;; start-res) ;; return the server info
;; (if (member remaining-tries '(3 4 6))
;; (begin ;; login failed
;; (debug:print 25 "INFO: client:setup start-res=" start-res ", run-id=" run-id ", server-dat=" host-info)
;; (hash-table-delete! *runremote* run-id)
;; (open-run-close tasks:server-force-clean-run-record
;; tasks:open-db
;; run-id
;; (car host-info)
;; (cadr host-info)
;; " client:setup (host-info=#t)")
;; (thread-sleep! 5)
;; (client:setup run-id remaining-tries: 10)) ;; (- remaining-tries 1)))
;; (begin
;; (debug:print 25 "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)))
;; (debug:print-info 0 "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)))
;; (if start-res
;; (begin
;; (hash-table-set! *runremote* run-id start-res)
;; start-res)
;; (if (member remaining-tries '(2 5))
;; (begin ;; login failed
;; (debug:print 25 "INFO: client:setup start-res=" start-res ", run-id=" run-id ", server-dat=" server-dat)
;; (hash-table-delete! *runremote* run-id)
;; (open-run-close tasks:server-force-clean-run-record
;; tasks:open-db
;; run-id
;; (tasks:hostinfo-get-interface server-dat)
;; (tasks:hostinfo-get-port server-dat)
;; " client:setup (server-dat = #t)")
;; (thread-sleep! 2)
;; (server:try-running run-id)
;; (thread-sleep! 10) ;; give server a little time to start up
;; (client:setup run-id remaining-tries: 10)) ;; (- remaining-tries 1)))
;; (begin
;; (debug:print 25 "INFO: client:setup start-res=" start-res ", run-id=" run-id ", server-dat=" server-dat)
;; (thread-sleep! 5)
;; (client:setup run-id remaining-tries: (- remaining-tries 1))))))
;; (begin ;; no server registered
;; (if (eq? remaining-tries 2)
;; (begin
;; ;; (open-run-close tasks:server-clean-out-old-records-for-run-id tasks:open-db run-id " client:setup (server-dat=#f)")
;; (client:setup run-id remaining-tries: 10))
;; (begin
;; (thread-sleep! 2)
;; (debug:print 25 "INFO: client:setup start-res (not defined here), run-id=" run-id ", server-dat=" server-dat)
;; (if (< (open-run-close tasks:num-in-available-state tasks:open-db run-id) 3)
;; (begin
;; ;; (open-run-close tasks:server-clean-out-old-records-for-run-id tasks:open-db run-id " client:setup (server-dat=#f)")
;; (server:try-running run-id)))
;; (thread-sleep! 10) ;; give server a little time to start up
;; (client:setup run-id remaining-tries: (- remaining-tries 1)))))))))))
;; Do all the connection work, look up the transport type and set up the
;; connection if required.
;;
;; There are two scenarios.
;; 1. We are a test manager and we received *transport-type* and *runremote* via cmdline
;; 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 #!key (numtries 3))
(if (not *toppath*)
(define (client:setup-http run-id #!key (remaining-tries 10) (failed-connects 0))
(debug:print-info 2 "client:setup remaining-tries=" remaining-tries)
(let* ((tdbdat (tasks:open-db)))
(if (<= remaining-tries 0)
(if (not (launch:setup-for-run))
(begin
(debug:print 0 "ERROR: failed to find megatest.config, exiting")
(exit))))
(begin
(debug:print 0 "ERROR: failed to start or connect to server for run-id " run-id)
(exit 1))
(push-directory *toppath*) ;; This is probably NOT needed
(debug:print-info 11 "*transport-type* is " *transport-type* ", *runremote* is " *runremote*)
(let* ((hostinfo (open-run-close tasks:get-best-server tasks:open-db)))
(debug:print-info 11 "CLIENT SETUP, hostinfo=" hostinfo)
(set! *transport-type* (if hostinfo
(string->symbol (tasks:hostinfo-get-transport hostinfo))
'fs))
(debug:print-info 11 "Using transport type of " *transport-type* (if hostinfo (conc " to connect to " hostinfo) ""))
(case *transport-type*
(let* ((server-dat (tasks:get-server (db:delay-if-busy tdbdat) run-id)))
(debug:print-info 4 "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 "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 "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 "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)))))))))
;; (let ((host-info (hash-table-ref/default *runremote* run-id #f)))
;; (if host-info ;; this is a bit circular. the host-info *is* the start-res FIXME
;; (let* ((iface (http-transport:server-dat-get-iface host-info))
;; (port (http-transport:server-dat-get-port host-info))
;; (start-res (case *transport-type*
((fs)(if (not *megatest-db*)(set! *megatest-db* (open-db))))
((http)
(http-transport:client-connect (tasks:hostinfo-get-interface hostinfo)
(tasks:hostinfo-get-port hostinfo)))
((zmq)
(zmq-transport:client-connect (tasks:hostinfo-get-interface hostinfo)
(tasks:hostinfo-get-port hostinfo)
;; ((http)(http-transport:client-connect iface port))
;; ((nmsg)(nmsg-transport:client-connect iface port)) ;; (http-transport:server-dat-get-socket host-info))
;; (else #f)))
;; (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
;; (vector-ref (vector-ref logininfo 1) 1)
;; #f)))
;; (else #f))))
;; (if ping-res ;; sucessful login?
;; (begin
;; (debug:print-info 2 "client:setup, ping is good using host-info=" host-info ", remaining-tries=" remaining-tries)
;; start-res) ;; return the server info
;; ;; have host info but no ping. shutdown the current connection and try again
;; (begin ;; login failed
;; (debug:print-info 1 "client:setup, ping is bad for start-res=" start-res " and *runremote*=" host-info)
;; (case *transport-type*
;; ((http)(http-transport:close-connections run-id)))
;; (hash-table-delete! *runremote* run-id)
;; (if (< remaining-tries 8)
;; (thread-sleep! 5)
;; (thread-sleep! 1))
;; (client:setup run-id remaining-tries: (- remaining-tries 1)))))
;; ;; YUK: rename server-dat here
;;
;; 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)))
(tasks:hostinfo-get-pubport hostinfo)))
(else ;; default to fs
(debug:print 0 "ERROR: unrecognised transport type " *transport-type* " attempting to continue with fs")
(set! *transport-type* 'fs)
(set! *megatest-db* (open-db))))
(pop-directory)))
;; client:signal-handler
(define (client:signal-handler signum)
(signal-mask! signum)
(handle-exceptions
exn
(debug:print " ... exiting ...")
(let ((th1 (make-thread (lambda ()
"") ;; do nothing for now (was 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.")
(thread-sleep! 1) ;; give the flush one second to do it's stuff
(debug:print 0 " Done.")
(exit 4))
"exit on ^C timer")))
(thread-start! th2)
(thread-start! th1)
(thread-join! th2))))
;; client:launch
;; Need to set the signal handler somewhere other than here as this
;; routine will go away.
;;
(define (client:launch)
(define (client:launch run-id)
(set-signal-handler! signal/int client:signal-handler)
(if (client:setup)
(debug:print-info 2 "connected as client")
(begin
(debug:print 0 "ERROR: Failed to connect as client")
(exit))))
(if (client:setup run-id)
(debug:print-info 2 "connected as client")
(begin
(debug:print 0 "ERROR: Failed to connect as client")
(exit))))
|