Megatest

;; 
;; (use spiffy uri-common intarweb http-client spiffy-request-vars intarweb spiffy-directory-listing)
;; 
;; Configurations for server
(tcp-buffer-size 2048)
(max-connections 2048) 

(defstruct servdat
  host
  port
  uuid)

(define (http-transport:make-server-url hostport)
  (if (not hostport)
      #f

		
		;; get_next_port goes here
		(http-transport:try-start-server ipaddrstr
						 (portlogger:open-run-close portlogger:find-port)))
	      (begin
		(print "ERROR: Tried and tried but could not start the server"))))
      ;; any error in following steps will result in a retry
      (set! *server-info* (make-servdat host: ipaddrstr port: portnum))
      (debug:print 0 *default-log-port* "INFO: Trying to start server on " ipaddrstr ":" portnum)
      ;; This starts the spiffy server
      ;; NEED WAY TO SET IP TO #f TO BIND ALL
      ;; (start-server bind-address: ipaddrstr port: portnum)
      (if config-hostname ;; this is a hint to bind directly
	  (start-server port: portnum bind-address: (if (equal? config-hostname "-")
							ipaddrstr

	 (api-req      (make-request method: 'POST uri: api-uri))
	 (server-dat   (vector iface port api-uri api-url api-req (current-seconds) server-id)))
    server-dat))

;;======================================================================
;; NEW SERVER METHOD
;;======================================================================

;; only use for main.db - need to re-write some of this :(
;;
(define (get-lock-db sdat dbfile)
  (let* ((dbh (db:open-run-db dbfile db:initialize-db))
	 (res (db:get-iam-server-lock dbh dbfile)))
    (sqlite3:finalize! dbh)
    res))


(define *srvpktspec*
  `((server (host    . h)
	    (port    . p)
	    (servkey . k)
	    (pid     . i)
	    (ipaddr  . a)

		   pkts-dir
		   pkt-dat
		   pktspec: pkt-spec
		   ptype: 'server)))
    (debug:print 0 *default-log-port* "Server on "host":"port" registered in pkt "uuid)
    uuid))






(define (get-pkts-dir #!optional (apath #f))
  (let* ((effective-toppath (or *toppath* apath)))
    (assert effective-toppath
	    "ERROR: get-pkts-dir called without *toppath* set. Exiting.")
    (let* ((pdir (conc effective-toppath "/.meta/srvpkts")))
      (if (file-exists? pdir)
	  pdir

				;; create a server pkt in *toppath*/.meta/srvpkts

				;; TODO:
				;;   1. change sdat to stuct
				;;   2. add uuid to struct
				;;   3. update uuid in sdat here
				;;
				(servdat-uuid-set! sdat
						(register-server
						 pkts-dir *srvpktspec*
						 (get-host-name)
						 (servdat-port sdat) server-key
						 (servdat-host sdat) db-file))

				;; now read pkts and see if we are a contender
				(let* ((all-pkts     (get-all-server-pkts pkts-dir *srvpktspec*))
				       (viables      (get-viable-servers all-pkts db-file))
				       (best-srv     (get-best-candidate viables db-file))
				       (best-srv-key (if best-srv (alist-ref 'servkey best-srv) #f)))
				  (debug:print 0 *default-log-port* "best-srv-key: "best-srv-key", server-key: "server-key)
				  ;; am I the best-srv, compare server-keys to know
				  (if (equal? best-srv-key server-key)

				      (if (get-lock-db sdat db-file) ;; (db:get-iam-server-lock *dbstruct-db* *toppath* run-id)
					  (debug:print 0 *default-log-port* "I'm the server!")
					  (begin
					    (debug:print 0 *default-log-port* "I'm not the server, exiting.")
					    (bdat-time-to-exit-set! *bdat* #t)
					    (exit)))
				      (begin
					(debug:print 0 *default-log-port* "Keys do not match "best-srv-key", "server-key", exiting.")
					(bdat-time-to-exit-set! *bdat* #t)))
				  sdat))
                              (begin
				(debug:print-info 0 *default-log-port* "Still waiting, last-sdat=" last-sdat)
                                (sleep 4)
				(if (> (- (current-seconds) start-time) 120) ;; been waiting for two minutes
				    (begin
				      (debug:print-error 0 *default-log-port* "transport appears to have died, exiting server")
                                      #;(common:save-pkt `((action . died)
                                                         (T      . server)
                                                         (pid    . ,(current-process-id))
                                                         (ipaddr . ,(car sdat))
                                                         (port   . ,(cadr sdat))
                                                         (msg    . "Transport died?"))
						       *configdat* #t)
				      (exit))
				    (loop start-time
					  (equal? sdat last-sdat)
					  sdat)))))))
	 (iface       (servdat-host server-info))
         (port        (servdat-port server-info))
         (last-access 0)
	 (server-timeout (server:expiration-timeout))
	 (server-log-file (args:get-arg "-log"))) ;; always set when we are a server

    (let loop ((count         0)
	       (server-state 'available)
	       (bad-sync-count 0)
	       (start-time     (current-milliseconds)))
      ;; Use this opportunity to sync the tmp db to megatest.db
      (if (not *dbstruct-db* )
	  (let ((watchdog (bdat-watchdog *bdat*)))
	    (debug:print 0 *default-log-port* "SERVER: dbprep")
	    
	    (db:setup dbname) ;; sets *dbstruct-db* as side effect

	    (debug:print 0 *default-log-port* "SERVER: running, megatest version: " (common:get-full-version)) ;; NOTE: the server is NOT yet marked as running in the log. We do that in the keep-running routine.
	    (if watchdog
		(if (not (member (thread-state watchdog) '(ready running blocked sleeping dead)))
		    (begin
		      (debug:print-info 0 *default-log-port* "Starting watchdog thread (in state "(thread-state watchdog)")")
		      (thread-start! watchdog)))
		(debug:print 0 *default-log-port* "ERROR: *watchdog* not setup, cannot start it."))))
      
      ;; when things go wrong we don't want to be doing the various queries too often
      ;; so we strive to run this stuff only every four seconds or so.
      (let* ((sync-time (- (current-milliseconds) start-time))
	    (rem-time  (quotient (- 4000 sync-time) 1000)))
	(if (and (<= rem-time 4)
		 (>  rem-time 0))
	    (thread-sleep! rem-time)))
      
      (if (< count 1) ;; 3x3 = 9 secs aprox
	  (loop (+ count 1) 'running bad-sync-count (current-milliseconds)))
      
      ;; Check that iface and port have not changed (can happen if server port collides)
      (mutex-lock! *heartbeat-mutex*)
      (set! sdat *server-info*)
      (mutex-unlock! *heartbeat-mutex*)
      
      (if (or (not (equal? (servdat-host sdat) iface))
	      (not (equal? (servdat-port sdat) port)))
	  (let ((new-iface (servdat-host sdat))
		(new-port  (servdat-port sdat)))
	    (debug:print-info 0 *default-log-port* "WARNING: interface changed, refreshing iface and port info")
	    (set! iface new-iface)
	    (set! port  new-port)
	    (if (not *server-id*)
		(set! *server-id* (server:mk-signature)))
            ;; (debug:print 0 *default-log-port* (current-seconds) (current-directory) (current-process-id) (argv))
	    (debug:print 0 *default-log-port* "SERVER STARTED: " iface ":" port " AT " (current-seconds) " server-id: " *server-id*)
	    (flush-output *default-log-port*)))
      
      ;; Transfer *db-last-access* to last-access to use in checking that we are still alive
      (mutex-lock! *heartbeat-mutex*)
      (set! last-access *db-last-access*)
      (mutex-unlock! *heartbeat-mutex*)
      
      (if (common:low-noise-print 120 (conc "server running on " iface ":" port))
	  (begin
             (if (not *server-id*)
              (set! *server-id* (server:mk-signature)))
	     ;; (debug:print 0 *default-log-port* (current-seconds) (current-directory) (current-process-id) (argv))   
	     (debug:print 0 *default-log-port* "SERVER STARTED: " iface ":" port " AT " (current-seconds) " server-id: " *server-id*)
	     (flush-output *default-log-port*)))
      (if (common:low-noise-print 60 "dbstats")
	  (begin
	    (debug:print 0 *default-log-port* "Server stats:")
	    (db:print-current-query-stats)))
      (let* ((hrs-since-start  (/ (- (current-seconds) server-start-time) 3600)))
	(cond
         ((and *server-run*