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
|
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
|
-
+
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
-
-
-
-
-
-
-
+
+
+
+
+
+
-
-
+
-
-
-
-
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
|
;; 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 run-id #!key (remaining-tries 100) (failed-connects 0))
(define (client:setup run-id #!key (remaining-tries 2) (failed-connects 0))
(debug:print-info 2 "client:setup remaining-tries=" remaining-tries)
(let* ((tdbdat (tasks:open-db)))
(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)))
(if host-info
(let* ((iface (http-transport:server-dat-get-iface host-info))
(port (http-transport:server-dat-get-port host-info))
(start-res (http-transport:client-connect iface port))
(ping-res (rmt:login-no-auto-client-setup start-res run-id)))
(if ping-res ;; sucessful login?
(begin
(debug:print-info 2 "client:setup, ping is good using host-info=" host-info ", remaining-tries=" remaining-tries)
;; Why add the close-connections here?
;; (http-transport:close-connections run-id)
(hash-table-set! *runremote* run-id start-res)
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)
(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
(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))
(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
(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)
(http-transport:close-connections run-id)
(hash-table-delete! *runremote* 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)")
(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: (- 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)
(thread-sleep! 2)
(if (< num-available 2)
(begin
(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)))))))))))
|