Megatest

     (debug:print 0 *default-log-port* " message: "  ((condition-property-accessor 'exn 'message) exn))       
     (vector #f (vector exn call-chain dat))) ;; return some stuff for debug if an exception happens
   (cond
    ((not (vector? dat))                    ;; it is an error to not receive a vector
     (vector #f (vector #f "remote must be called with a vector")))
    ((> *api-process-request-count* 20) ;; 20)
     (debug:print 0 *default-log-port* "WARNING: api:execute-requests received an overloaded message.")
     (vector #f (vector 'overloaded))) ;; the inner vector is what gets returned. nope, don't know why. please refactor!
     (vector #f (vector #f 'overloaded))) ;; the inner vector is what gets returned. nope, don't know why. please refactor!
    (else  
     (let* ((cmd-in            (vector-ref dat 0))
            (cmd               (if (symbol? cmd-in)
				   cmd-in
				   (string->symbol cmd-in)))
            (params            (vector-ref dat 1))
            (start-t           (current-milliseconds))

       ;; send the data and get the response
       ;; extract the needed info from the http data and 
       ;; process and return it.
       (let* ((send-recieve (lambda ()
			      (mutex-lock! *http-mutex*)
			      ;; (condition-case (with-input-from-request "http://localhost"; #f read-lines)
			      ;;					       ((exn http client-error) e (print e)))
			      (set! res (vector
			      (set! res (vector                ;;; DON'T FORGET - THIS IS THE CLIENT SIDE! NOTE: consider moving this to client.scm since we are only supporting http transport at this time.
					 success
					 (db:string->obj 
					  (handle-exceptions
					      exn
					      (begin
					      (let ((call-chain (get-call-chain))
						    (msg        ((condition-property-accessor 'exn 'message) exn)))
						(set! success #f)
						(debug:print 0 *default-log-port* "WARNING: failure in with-input-from-request to " fullurl ".")
						(debug:print 0 *default-log-port* " message: " ((condition-property-accessor 'exn 'message) exn))
						(debug:print 0 *default-log-port* " message: " msg)
						(debug:print 0 *default-log-port* " cmd: " cmd " params: " params)
						(if runremote
						    (remote-conndat-set! runremote #f))
						;; Killing associated server to allow clean retry.")
						;; (tasks:kill-server-run-id run-id)  ;; better to kill the server in the logic that called this routine?
						(mutex-unlock! *http-mutex*)
					     ;;; (signal (make-composite-condition
					     ;;;          (make-property-condition 'commfail 'message "failed to connect to server")))

		     (begin
		       (debug:print-info 0 *default-log-port* "Attempting to kill server with pid " pid)
		       (server:kill server)))))
	     (sort servers (lambda (a b)
			     (let ((ma (or (any->number (car a)) 9e9))
				   (mb (or (any->number (car b)) 9e9)))
			       (> ma mb)))))
	    (debug:print-info 1 *default-log-port* "Done with listservers")
	    ;; (debug:print-info 1 *default-log-port* "Done with listservers")
	    (set! *didsomething* #t)
	    (exit))
	  (exit))))
      ;; must do, would have to add checks to many/all calls below

;;======================================================================
;; Weird special calls that need to run *after* the server has started?

	     (success  (if (vector? dat) (vector-ref dat 0) #f))
	     (res      (if (vector? dat) (vector-ref dat 1) #f)))
	(if (vector? conninfo)(http-transport:server-dat-update-last-access conninfo)) ;; refresh access time
	;; (mutex-unlock! *rmt-mutex*)
        (debug:print-info 13 *default-log-port* "rmt:send-receive, case  9. conninfo=" conninfo " dat=" dat " runremote = " runremote)
	(mutex-unlock! *rmt-mutex*)
	(if success ;; success only tells us that the transport was successful, have to examine the data to see if there was a detected issue at the other end
	    (if (and (symbol? res)
		     (eq? res 'overloaded))
	    (if (and (vector? res)
		     (eq? (vector-length res) 2)
		     (eq? (vector-ref res 2) 'overloaded)) ;; since we are looking at the data to carry the error we'll use a fairly obtuse combo to minimise the chances of some sort of collision.
		(let ((wait-delay (+ attemptnum (* attemptnum 10))))
		  (debug:print 0 *default-log-port* "WARNING: server is overloaded. Delaying " wait-delay " seconds and trying call again.")
		  (mutex-lock! *rmt-mutex*)
		  (set! *runremote* #f) ;; force starting over
		  (mutex-unlock! *rmt-mutex*)
		  (thread-sleep! wait-delay)
		  (rmt:send-receive cmd rid params attemptnum: (+ attemptnum 1)))
		res) ;; All good, return res
	    (begin
	      (debug:print 0 *default-log-port* "WARNING: communication failed. Trying again, try num: " attemptnum)
	      (remote-conndat-set!    runremote #f)
	      (remote-server-url-set! runremote #f)

					       (exit 4)))))
		       (thread-start! th2)
		       (thread-start! th1)
		       (thread-join! th2)))))
      (set-signal-handler! signal/int sighand)
      (set-signal-handler! signal/term sighand))

    ;; force the starting of a server
    (debug:print 0 *default-log-port* "waiting on server...")
    (server:start-and-wait *toppath*)
    
    (runs:set-megatest-env-vars run-id inkeys: keys inrunname: runname) ;; these may be needed by the launching process
    (set! runconf (if (file-exists? runconfigf)
		      (setup-env-defaults runconfigf run-id *already-seen-runconfig-info* keyvals target)
		      (begin
			(debug:print 0 *default-log-port* "WARNING: You do not have a run config file: " runconfigf)
			#f)))

			   waitons:     waitons
			   testmode:    testmode
			   newtal:      newtal
			   itemmaps:    itemmaps
			   ;; prereqs-not-met: prereqs-not-met
			   )))
	(runs:dat-regfull-set! runsdat regfull)

	;; every couple minutes verify the server is there for this run
	;; (if (and (common:low-noise-print 60 "try start server"  run-id)
	;; 	 (tasks:need-server run-id))
	;; every 15 minutes verify the server is there for this run
	(if (and (common:low-noise-print 240 "try start server"  run-id)
		 (not (server:check-if-running *toppath*)))
	    (server:kind-run *toppath*))
	;;     (tasks:start-and-wait-for-server tdbdat run-id 10)) ;; NOTE: delay and wait is done under the hood
	
	(if (> num-running 0)
	  (set! last-time-some-running (current-seconds)))

      (if (> (current-seconds)(+ last-time-some-running (or (configf:lookup *configdat* "setup" "give-up-waiting") 36000)))
	  (hash-table-set! *max-tries-hash* tfullname (+ (hash-table-ref/default *max-tries-hash* tfullname 0) 1)))
	;; (debug:print 0 *default-log-port* "max-tries-hash: " (hash-table->alist *max-tries-hash*))

;;
;; mod-time host port start-time pid
;;
;; sort by start-time descending. I.e. get the oldest first. Young servers will thus drop off
;; and servers should stick around for about two hours or so.
;;
(define (server:get-best srvlst)
  (let* ((nums (server:get-num-servers))
  (let ((now (current-seconds)))
    (sort
     (filter (lambda (rec)
	       (if (and (list? rec)
			(> (length rec) 2))
		   (let ((start-time (list-ref rec 3))
			 (mod-time   (list-ref rec 0)))
		     ;; (print "start-time: " start-time " mod-time: " mod-time)
		     (and start-time mod-time
			  (> (- now start-time) 0)    ;; been running at least 0 seconds
			  (< (- now mod-time)   16)   ;; still alive - file touched in last 16 seconds
			  (< (- now start-time) 
			     (+ (- (string->number (or (configf:lookup *configdat* "server" "runtime") "3600"))
				   180)
				(random 360))) ;; under one hour running time +/- 180
			  ))
		   #f))
	     srvlst)
     (lambda (a b)
       (< (list-ref a 3)
	  (list-ref b 3))))))
	 (now  (current-seconds))
	 (slst (sort
		(filter (lambda (rec)
			  (if (and (list? rec)
				   (> (length rec) 2))
			      (let ((start-time (list-ref rec 3))
				    (mod-time   (list-ref rec 0)))
				;; (print "start-time: " start-time " mod-time: " mod-time)
				(and start-time mod-time
				     (> (- now start-time) 0)    ;; been running at least 0 seconds
				     (< (- now mod-time)   16)   ;; still alive - file touched in last 16 seconds
				     (< (- now start-time) 
					(+ (- (string->number (or (configf:lookup *configdat* "server" "runtime") "3600"))
					      180)
					   (random 360))) ;; under one hour running time +/- 180
				     ))
			      #f))
			srvlst)
		(lambda (a b)
		  (< (list-ref a 3)
		     (list-ref b 3))))))
    (if (> (length slst) nums)
	(take slst nums)
	slst)))

(define (server:get-first-best areapath)
  (let ((srvrs (server:get-best (server:get-list areapath))))
    (if (and srvrs
	     (not (null? srvrs)))
	(car srvrs)
	#f)))

	  (let ((num-ok (length (server:get-best (server:get-list areapath)))))
	    (if (< num-ok 1) ;; if there are no decent candidates for servers then try starting a new one
		(server:kind-run areapath))
	    (thread-sleep! 5)
	    (loop (server:check-if-running areapath)))))))

(define server:try-running server:run) ;; there is no more per-run servers ;; REMOVE ME. BUG.

(define (server:get-num-servers #!key (numservers 2))
  (let ((ns (string->number
	     (or (configf:lookup *configdat* "server" "numservers") "notanumber"))))
    (or ns numservers)))

;; no longer care if multiple servers are started by accident. older servers will drop off in time.
;;
(define (server:check-if-running areapath #!key (numservers "2"))
  (let* ((ns            (string->number
(define (server:check-if-running areapath) ;;  #!key (numservers "2"))
  (let* ((ns            (server:get-num-servers))
			 (or (configf:lookup *configdat* "server" "numservers") numservers)))
	 (servers       (server:get-best (server:get-list areapath))))
    ;; (print "servers: " servers " ns: " ns)
    (if (or (and servers
		 (null? servers))
	    (not servers)
	    (and (list? servers)
		 (< (length servers) (random ns)))) ;; somewhere between 0 and numservers