Megatest

;;    - keep it simple, only return the actual result of the call, i.e. no meta info here
;;
;;    - returns #( flag result )
;;
(define (api:execute-requests dbstruct dat)
  (handle-exceptions
   exn
   (let ((call-chain (get-call-chain)))

     (debug:print 0 *default-log-port* "WARNING: api:execute-requests received an exception from peer")
     (print-call-chain (current-error-port))
     (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)
     '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))
            (readonly-mode     (dbr:dbstruct-read-only dbstruct))
            (readonly-command  (member cmd api:read-only-queries))
            (writecmd-in-readonly-mode (and readonly-mode (not readonly-command)))
            (res    
             (if writecmd-in-readonly-mode
                 (conc "attempt to run write command "cmd" on a read-only database")
                 (case cmd
                   ;;===============================================
                   ;; READ/WRITE QUERIES

                   ((sdb-qry)                      (apply sdb:qry params))
                   ((ping)                         (current-process-id))

                   ;; TESTMETA
                   ((testmeta-get-record)       (apply db:testmeta-get-record dbstruct params))

                   ;; TASKS 
                   ((find-task-queue-records)   (apply tasks:find-task-queue-records dbstruct params))
		   (else
		    (debug:print 0 *default-log-port* "ERROR: bad api call " cmd)
		    (conc "ERROR: BAD api call " cmd))))))
       
       ;; save all stats
       (let ((delta-t (- (current-milliseconds)
			 start-t)))
	 (hash-table-set! *db-api-call-time* cmd
			  (cons delta-t (hash-table-ref/default *db-api-call-time* cmd '()))))
       (if writecmd-in-readonly-mode
	   (vector #f res)
           (vector #t res)))))))

;; http-server  send-response
;;                 api:process-request
;;                    db:*
;;
;; NB// Runs on the server as part of the server loop
;;
(define (api:process-request dbstruct $) ;; the $ is the request vars proc
  (set! *api-process-request-count* (+ *api-process-request-count* 1))
  (let* ((cmd     ($ 'cmd))
	 (paramsj ($ 'params))
	 (params  (db:string->obj paramsj transport: 'http)) ;; incoming data from the POST (or is it a GET?)
	 (resdat  (api:execute-requests dbstruct (vector cmd params))) ;; process the request, resdat = #( flag result )
	 (success (vector-ref resdat 0))
	 (res     (vector-ref resdat 1))) ;; (vector flag payload), get the payload, ignore the flag (why?)
    (if (not success)
	(debug:print 0 *default-log-port* "ERROR: success flag is #f for " cmd " with params " params))
    (if (> *api-process-request-count* *max-api-process-requests*)
	(set! *max-api-process-requests* *api-process-request-count*))
    (set! *api-process-request-count* (- *api-process-request-count* 1))
    ;; This can be here but needs controls to ensure it doesn't run more than every 4 seconds
    ;; (rmt:dat->json-str
    ;;  (if (or (string? res)
    ;;          (list?   res)
    ;;          (number? res)
    ;;          (boolean? res))
    ;;      res 
    ;;      (list "ERROR, not string, list, number or boolean" 1 cmd params res)))))
    (db:obj->string res transport: 'http)))

      (begin
	(debug:print-error 0 *default-log-port* "failed to start or connect to server")
	(exit 1))
      ;;
      ;; Alternatively here, we can get the list of candidate servers and work our way
      ;; through them searching for a good one.
      ;;
      (let* ((server-dat (server:get-rand-best areapath)) ;; (server:get-first-best areapath))
	     (runremote  (or area-dat *runremote*)))
	(if (not server-dat) ;; no server found
	    (client:setup-http areapath remaining-tries: (- remaining-tries 1))
	    (let ((host  (cadr  server-dat))
		  (port  (caddr server-dat)))
	      (debug:print-info 4 *default-log-port* "client:setup server-dat=" server-dat ", remaining-tries=" remaining-tries)
	      (if (and (not area-dat)

(defstruct remote
  (hh-dat            (common:get-homehost)) ;; homehost record ( addr . hhflag )
  (server-url        (if *toppath* (server:check-if-running *toppath*))) ;; (server:check-if-running *toppath*) #f))
  (last-server-check 0)  ;; last time we checked to see if the server was alive
  (conndat           #f)
  (transport         *transport-type*)
  (server-timeout    (or (server:get-timeout) 100)) ;; default to 100 seconds
  (force-server      #f)
  (ro-mode           #f)  
  (ro-mode-checked   #f)) ;; flag that indicates we have checked for ro-mode

;; launching and hosts
(defstruct host
  (reachable    #f)
  (last-update  0)
  (last-used    0)
  (last-cpuload 1))

   (read (open-input-string (z3:decode-buffer (base64:base64-decode instr))))))

;; dot-locking egg seems not to work, using this for now
;; if lock is older than expire-time then remove it and try again
;; to get the lock
;;
(define (common:simple-file-lock fname #!key (expire-time 300))
  (handle-exceptions
      exn
      #f ;; don't really care what went wrong right now. NOTE: I have not seen this one actually fail.
    (if (file-exists? fname)
	(if (> (- (current-seconds)(file-modification-time fname)) expire-time)
	    (begin
	      (delete-file* fname)
	      (common:simple-file-lock fname expire-time: expire-time))
	    #f)
	(let ((key-string (conc (get-host-name) "-" (current-process-id))))
	  (with-output-to-file fname
	    (lambda ()
	      (print key-string)))
	  (thread-sleep! 0.25)
	  (if (file-exists? fname)
	      (with-input-from-file fname
		(lambda ()
		  (equal? key-string (read-line))))
	      #f)))))

(define (common:simple-file-lock-and-wait fname #!key (expire-time 300))
  (let ((end-time (+ expire-time (current-seconds))))
    (let loop ((got-lock (common:simple-file-lock fname expire-time: expire-time)))
      (if got-lock
	  #t
	  (if (> end-time (current-seconds))
	      (begin
		(thread-sleep! 3)
		(loop (common:simple-file-lock fname expire-time: expire-time)))
	      #f)))))

(define (common:simple-file-release-lock fname)
  (handle-exceptions
      exn
      #f ;; I don't really care why this failed (at least for now)
    (delete-file* fname)))

;;======================================================================
;; S T A T E S   A N D   S T A T U S E S
;;======================================================================

(define *common:std-states*   
  '((0 "ARCHIVED")

;; Force creation of the db in case it isn't already there.
(tasks:open-db)

(define (dashboard:get-youngest-run-db-mod-time dbdir)
  (handle-exceptions
   exn
   (begin
     (debug:print 2 *default-log-port* "WARNING: error in accessing databases in get-youngest-run-db-mod-time: " ((condition-property-accessor 'exn 'message) exn) " db-dir="dbdir)
     (current-seconds)) ;; something went wrong - just print an error and return current-seconds
   (common:max (map (lambda (filen)
		      (file-modification-time filen))
		    (glob (conc dbdir "/*.db*"))))))

(define (dashboard:monitor-changed? commondat tabdat)
  (let* ((run-update-time (current-seconds))

(define (db:print-current-query-stats)
  ;; generate stats from *db-api-call-time*
  (let ((ordered-keys (sort (hash-table-keys *db-api-call-time*)
			    (lambda (a b)
			      (let ((sum-a (common:sum (hash-table-ref *db-api-call-time* a)))
				    (sum-b (common:sum (hash-table-ref *db-api-call-time* b))))
				(> sum-a sum-b)))))
	(total        0))
    (for-each
     (lambda (cmd-key)
       (let* ((dat  (hash-table-ref *db-api-call-time* cmd-key))
	      (num  (length dat))
	      (avg  (if (> num 0)
			(/ (common:sum dat)(length dat)))))
	 (set! total (+ total num))
	 (debug:print-info 0 *default-log-port* cmd-key "\tavg: " avg " max: " (common:max dat) " min: " (common:min-max < dat) " num: " (length dat))))
     ordered-keys)
    (debug:print-info 0 *default-log-port* "TOTAL: " total " api calls since start.")))

(define (db:get-all-run-ids dbstruct)
  (db:with-db
   dbstruct
   #f
   #f
   (lambda (db)

	      (th2 (make-thread time-out     "time out")))
	 (thread-start! th1)
	 (thread-start! th2)
	 (thread-join! th1)
	 (thread-terminate! th2)
	 (debug:print-info 11 *default-log-port* "got res=" res)
	 (if (vector? res)
	     (if (vector-ref res 0) ;; this is the first flag or the second flag?
		 res ;; this is the *inner* vector? seriously? why?
                 (if (debug:debug-mode 11)
                     (begin ;; note: this code also called in nmsg-transport - consider consolidating it
                       (debug:print-error 11 *default-log-port* "error occured at server, info=" (vector-ref res 2))
                       (debug:print 11 *default-log-port* " client call chain:")
                       (print-call-chain (current-error-port))
                       (debug:print 11 *default-log-port* " server call chain:")
                       (pp (vector-ref res 1) (current-error-port))

      (set! last-access *db-last-access*)
      (mutex-unlock! *heartbeat-mutex*)
      
      (if (common:low-noise-print 120 (conc "server running on " iface ":" port))
	  (begin
	    (debug:print 0 *default-log-port* "SERVER STARTED: " iface ":" port " AT " (current-seconds))
	    (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))
	     (adjusted-timeout (if (> hrs-since-start 1)
				   (- server-timeout (inexact->exact (round (* hrs-since-start 60))))  ;; subtract 60 seconds per hour
				   server-timeout)))
	(if (common:low-noise-print 120 "server timeout")
	    (debug:print-info 0 *default-log-port* "Adjusted server timeout: " adjusted-timeout))
	(cond

                                 (start-res (http-transport:client-connect host port))
                                 (ping-res  (rmt:login-no-auto-client-setup start-res)))
			    (if (and start-res
				     ping-res)
				(let ((url  (http-transport:server-dat-make-url start-res)))
				  (remote-conndat-set! *runremote* start-res)
				  (remote-server-url-set! *runremote* url)
				  (if (server:ping url)
				      (debug:print-info 0 *default-log-port* "connected to " url " using CMDINFO data.")
				      (begin
					(debug:print-info 0 *default-log-port* "have CMDINFO data but failed to connect to " url)
					(remote-conndat-set! *runremote* #f)
					(remote-server-url-set! *runremote* #f))))
				(debug:print-info 0 *default-log-port* "received " host ":" port " for url but could not connect.")
				)))))))
	  ;; NFS might not have propagated the directory meta data to the run host - give it time if needed
	  (let loop ((count 0))
	    (if (or (file-exists? top-path)
		    (> count 10))
		(change-directory top-path)

	     (target   (common:args-get-target))
	     (linktree (common:get-linktree))
	     (sections (if target (list "default" target) #f)) ;; for runconfigs
	     (mtconfig (or (args:get-arg "-config") "megatest.config")) ;; allow overriding megatest.config 
	     (rundir   (if (and runname target linktree)(conc linktree "/" target "/" runname) #f))
	     (mtcachef (and rundir (conc rundir "/" ".megatest.cfg-"  megatest-version "-" megatest-fossil-hash)))
	     (rccachef (and rundir (conc rundir "/" ".runconfigs.cfg-"  megatest-version "-" megatest-fossil-hash)))
	     (cancreate (and rundir (common:file-exists? rundir)(file-write-access? rundir)))
	     (cxt       (hash-table-ref/default *contexts* toppath #f)))

	;; create our cxt for this area if it doesn't already exist
	(if (not cxt)(hash-table-set! *contexts* toppath (make-cxt)))

	;; (print "runname: " runname " target: " target " mtcachef: " mtcachef " rccachef: " rccachef)
	(set! *toppath* toppath) ;; This is needed when we are running as a test using CMDINFO as a datasource

  ;; do all the prep locked under the rmt-mutex
  (mutex-lock! *rmt-mutex*)

  ;; 1. check if server is started IFF cmd is a write OR if we are not on the homehost, store in runremote
  ;; 2. check the age of the connections. refresh the connection if it is older than timeout-20 seconds.
  ;; 3. do the query, if on homehost use local access
  ;;
  (let* ((start-time    (current-seconds)) ;; snapshot time so all use cases get same value
         (areapath      *toppath*);; TODO - resolve from dbstruct to be compatible with multiple areas
	 (runremote     (or area-dat
			    *runremote*))
	 (readonly-mode (if (and runremote
				 (remote-ro-mode-checked runremote))
			    (remote-ro-mode runremote)
			    (let* ((dbfile  (conc *toppath* "/megatest.db"))
				   (ro-mode (not (file-write-access? dbfile)))) ;; TODO: use dbstruct or runremote to figure this out in future
			      (if runremote
				  (begin
				    (remote-ro-mode-set! runremote ro-mode)
				    (remote-ro-mode-checked-set! runremote #t)
				    ro-mode)
				  ro-mode)))))

    ;;(print "BB> readonly-mode is "readonly-mode" dbfile is "dbfile)
    (cond
     ;; give up if more than 15 attempts
     ((> attemptnum 15)
      (debug:print 0 *default-log-port* "ERROR: 15 tries to start/connect to server. Giving up.")
      (exit 1))

	   (let ((expire-time (+ (- start-time (remote-server-timeout runremote))(random 30)))) ;; add 30 seconds of noise so that not all running tests expire at the same time causing a storm of server starts
	     (< (http-transport:server-dat-get-last-access (remote-conndat runremote)) expire-time)))
      (debug:print-info 12 *default-log-port* "rmt:send-receive, case  8")
      (remote-conndat-set! runremote #f)
      (mutex-unlock! *rmt-mutex*)
      (rmt:send-receive cmd rid params attemptnum: attemptnum))
     ;; ensure we have a record for our connection for given area
     ((not runremote)                  ;; can remove this one. should never get here.         
      (set! *runremote* (make-remote)) ;; new runremote will come from this on next iteration
      (mutex-unlock! *rmt-mutex*)
      (debug:print-info 12 *default-log-port* "rmt:send-receive, case  1")
      (rmt:send-receive cmd rid params attemptnum: attemptnum))
     ;; ensure we have a homehost record
     ((not (pair? (remote-hh-dat runremote)))  ;; not on homehost
      (thread-sleep! 0.1) ;; since we shouldn't get here, delay a little

     ((or (and (remote-force-server runremote)              ;; we are forcing a server and don't yet have a connection to one
	       (not (remote-conndat runremote)))
	  (and (not (cdr (remote-hh-dat runremote)))        ;; not on a homehost 
	       (not (remote-conndat runremote))))           ;; and no connection
      (debug:print-info 12 *default-log-port* "rmt:send-receive, case  6  hh-dat: " (remote-hh-dat runremote) " conndat: " (remote-conndat runremote))
      (mutex-unlock! *rmt-mutex*)
      (if (not (server:check-if-running *toppath*)) ;; who knows, maybe one has started up?
	  (server:start-and-wait *toppath*))
      (remote-force-server-set! runremote (common:force-server?))
      (remote-conndat-set! runremote (rmt:get-connection-info *toppath*)) ;; calls client:setup which calls client:setup-http
      (rmt:send-receive cmd rid params attemptnum: attemptnum)) ;; TODO: add back-off timeout as
     ;; all set up if get this far, dispatch the query
     ((and (not (remote-force-server runremote))
	   (cdr (remote-hh-dat runremote))) ;; we are on homehost
      (mutex-unlock! *rmt-mutex*)

			 (else
			  (debug:print 0 *default-log-port* "ERROR: transport " (remote-transport runremote) " not supported")
			  (exit))))
	     (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))
		(let ((wait-delay (+ attemptnum (* attemptnum 10))))
		  (debug:print 0 *default-log-port* "WARNING: server is overloaded. Delaying " wait-delay " seconds and trying call again.")

		  (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)
	      (debug:print-info 12 *default-log-port* "rmt:send-receive, case  9.1")
	      (if (not (server:check-if-running *toppath*))
		  (server:start-and-wait *toppath*))
	      (rmt:send-receive cmd rid params attemptnum: (+ attemptnum 1)))))))))

;; (define (rmt:update-db-stats run-id rawcmd params duration)
;;   (mutex-lock! *db-stats-mutex*)
;;   (handle-exceptions
;;    exn
;;    (begin

    (thread-join! log-rotate)
    (pop-directory)))

;; given a path to a server log return: host port startseconds
;;
(define (server:logf-get-start-info logf)
  (let ((rx (regexp "^SERVER STARTED: (\\S+):(\\d+) AT ([\\d\\.]+)"))) ;; SERVER STARTED: host:port AT timesecs
    (handle-exceptions
	exn
	(list #f #f #f) ;; no idea what went wrong, call it a bad server
      (with-input-from-file
	  logf
	(lambda ()
	  (let loop ((inl  (read-line))
		     (lnum 0))
	    (if (not (eof-object? inl))
		(let ((mlst (string-match rx inl)))
		  (if (not mlst)
		      (if (< lnum 500) ;; give up if more than 500 lines of server log read
			  (loop (read-line)(+ lnum 1))
			  (list #f #f #f))
		      (let ((dat  (cdr mlst)))
			(list (car dat) ;; host
			      (string->number (cadr dat)) ;; port
			      (string->number (caddr dat))))))
		(list #f #f #f))))))))

;; get a list of servers with all relevant data
;; ( mod-time host port start-time pid )
;;
(define (server:get-list areapath #!key (limit #f))
  (let ((fname-rx    (regexp "^(|.*/)server-(\\d+)-(\\S+).log$"))
	(day-seconds (* 24 60 60)))
    ;; if the directory exists continue to get the list
    ;; otherwise attempt to create the logs dir and then
    ;; continue
    (if (if (directory-exists? (conc areapath "/logs"))

	    '()
	    (if (file-write-access? areapath)
		(begin
		  (condition-case
		      (create-directory (conc areapath "/logs") #t)
		    (exn (i/o file)(debug:print 0 *default-log-port* "ERROR: Cannot create directory at " (conc areapath "/logs")))
		    (exn ()(debug:print 0 *default-log-port* "ERROR: Unknown error attemtping to get server list.")))
		  (directory-exists? (conc areapath "/logs")))
		'()))
	(let* ((server-logs   (glob (conc areapath "/logs/server-*.log")))
	       (num-serv-logs (length server-logs)))
	  (if (null? server-logs)
	      '()
	      (let loop ((hed  (car server-logs))
			 (tal  (cdr server-logs))
			 (res '()))
		(let* ((mod-time  (handle-exceptions
				      exn

				      (current-seconds) ;; 0
				    (file-modification-time hed))) ;; default to *very* old so log gets ignored if deleted
		       (down-time (- (current-seconds) mod-time))
		       (serv-dat  (if (or (< num-serv-logs 10)
				  	  (< down-time day-seconds))
				     (server:logf-get-start-info hed)
				     '())) ;; don't waste time processing server files not touched in the past day if there are more than ten servers to look at
		       (serv-rec (cons mod-time serv-dat))
		       (fmatch   (string-match fname-rx hed))

;; 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 ((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))))))

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

(define (server:get-rand-best areapath)
  (let ((srvrs (server:get-best (server:get-list areapath))))
    (if (and (list? srvrs)
	     (not (null? srvrs)))
	(let* ((len (length srvrs))
	       (idx (random len)))
	  (list-ref srvrs idx))
	#f)))


(define (server:record->url servr)
  (match-let (((mod-time host port start-time pid)
	       servr))
    (if (and host port)
	(conc host ":" port)
	#f)))

	    (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.

;; 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
			 (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
        #f
        (let loop ((hed (car servers))
                   (tal (cdr servers)))
          (let ((res (server:check-server hed)))
            (if res
                res
                (if (null? tal)

		      (tests:generate-html-summary-for-iterated-test run-id test-id test-name outputfilename))
		  (common:simple-file-release-lock lockf)
		  (change-directory orig-dir)
		  ;; NB// tests:test-set-toplog! is remote internal...
		  (tests:test-set-toplog! run-id test-name outputfilename))
		;; didn't get the lock, check to see if current update started later than this 
		;; update, if so we can exit without doing any work
		(if (> my-start-time (handle-exceptions
					 exn
					 0
				       (file-modification-time lockf)))
		    ;; we started since current re-gen in flight, delay a little and try again
		    (begin
		      (debug:print-info 1 *default-log-port* "Waiting to update " outputfilename ", another test currently updating it")
		      (thread-sleep! (+ 5 (random 5))) ;; delay between 5 and 10 seconds
		      (loop (common:simple-file-lock lockf))))))))))

(define (tests:generate-html-summary-for-iterated-test run-id test-id test-name outputfilename)