Megatest

;;======================================================================
;; Copyright 2006-2012, Matthew Welland.
;; 
;; This file is part of Megatest.
;; 
;;     Megatest is free software: you can redistribute it and/or modify
;;     it under the terms of the GNU General Public License as published by
;;     the Free Software Foundation, either version 3 of the License, or

      (if (> num-tries 0) ;; should be "num-tries-left". 
	  (if (common:low-noise-print 30 (conc (round effective-load) "-load-acceptable-" effective-host))
	      (debug:print 0 *default-log-port* "Load on " effective-host " is acceptable at effective normalized load of "
			   effective-normalized-load "  continuing."))
	  (debug:print 0 *default-log-port* "Load on " effective-host ", "
		       first" could not be retrieved. Giving up and continuing."))))))

;; DO NOT CALL THIS DIRECTLY. It is called from common:wait-for-normalized-load
;;
;; (define (common:wait-for-cpuload maxload-in numcpus-in waitdelay #!key (count 1000) (msg #f)(remote-host #f)(force-maxload #f)(num-tries 5))
;;   (let* ((loadavg (common:get-cpu-load remote-host))
;; 	 (numcpus (if (<= 1 numcpus-in) ;; not possible to have zero.  If we get 1, it's possible that we got the previous default, and we should check again
;; 		      (common:get-num-cpus remote-host)
;; 		      numcpus-in))
;; 	 (maxload (if force-maxload
;; 		      maxload-in
;; 		      (if (number? maxload-in)
;; 			  (max maxload-in 0.5)
;; 			  0.5))) ;; so maxload must be greater than 0.5 for now BUG - FIXME?
;; 	 (first   (car loadavg))
;; 	 (next    (cadr loadavg))
;; 	 (adjmaxload (* maxload (max 1 numcpus))) ;; possible bug where
;; 					       ;; numcpus (or could be
;; 					       ;; maxload) is zero,
;; 					       ;; crude fallback is to
;; 					       ;; at least use 1
;; 	 (loadjmp (- first (if (> next (* numcpus 0.7)) ;; could do something with average of first and next?
;; 			       0
;; 			       next))) ;; we will force a conservative calculation any time next is large.
;; 	 (first-next-avg    (/ (+ first next) 2))
;; 	 ;; add some randomness to the time to break any alignment
;; 	 ;; where netbatch dumps many jobs to machines simultaneously
;;          (adjwait           (min (+ 300 (random 10)) (abs (* (+ (random 10)
;; 								(/ (- 1000 count) 10)
;; 								waitdelay)
;; 							     (- first adjmaxload) ))))
;; 	 (load-jump-limit   (configf:lookup-number *configdat* "setup" "load-jump-limit"))
;; 	 ;; effective load accounts for load jumps, this should elminate all the first-next-avg, adjwait, load-jump-limit
;; 	 ;; etc.
;; 	 (effective-load    (common:get-intercept first next))
;; 	 (effective-host    (or remote-host "localhost"))
;; 	 (normalized-effective-load (/ effective-load numcpus))
;; 	 (will-wait                 (> normalized-effective-load maxload)))
;; 	 
;;     ;; let's let the user know once in a long while that load checking
;;     ;; is happening but not constantly report it
;;     #;(if (common:low-noise-print 30 (conc "cpuload" (or remote-host "localhost"))) ;; (> (random 100) 75) ;; about 25% of the time
;; 	(debug:print-info 1 *default-log-port* "Checking cpuload on " (or remote-host "localhost") ", maxload: " maxload
;;     ", load: " first ", adjmaxload: " adjmaxload ", loadjmp: " loadjmp))
;; 
;;     (debug:print-info 1 *default-log-port*
;; 		      "On host: " effective-host
;; 		      ", effective load: " effective-load
;; 		      ", numcpus: " numcpus
;; 		      ", normalized effective load: " normalized-effective-load
;; 		      )
;;     
;;     (cond
;;      ;; bad data, try again to get the data
;;      ((and (< first 0) ;; this indicates the loadavg data is bad - machine may not be reachable
;; 	   (> num-tries 0))
;;       (debug:print 0 *default-log-port* "WARNING: received bad data from get-cpu-load " first ", we'll sleep 10s and try " num-tries " more times.")
;;       (thread-sleep! 10)
;;       (common:wait-for-cpuload maxload-in numcpus-in waitdelay
;; 			       count: count remote-host: remote-host force-maxload: force-maxload num-tries: (- num-tries 1)))
;;      ;; need to wait for load to drop
;;      ((and will-wait ;; (> first adjmaxload)
;; 	   (> count 0))
;;       (debug:print-info 0 *default-log-port*
;; 			"Delaying " 15 ;; adjwait
;; 			" seconds due to normalized effective load " normalized-effective-load ;; first
;; 			" exceeding max of " adjmaxload
;; 			" on server " (or remote-host (get-host-name))
;; 			" (normalized load-limit: " maxload ") " (if msg msg ""))
;;       (thread-sleep! 15) ;; adjwait)
;;       (common:wait-for-cpuload maxload numcpus waitdelay count: (- count 1) msg: msg remote-host: remote-host))
;;      ((and (> loadjmp (cond
;; 		       (load-jump-limit load-jump-limit)
;; 		       ((> numcpus 8)(/ numcpus 2))
;; 		       ((> numcpus 4)(/ numcpus 1.2))
;; 		       (else 0.5)))
;; 	   (> count 0))
;;       (debug:print-info 0 *default-log-port* "waiting " adjwait " seconds due to possible load jump " loadjmp ". "
;; 			(if msg msg ""))
;;       (thread-sleep! adjwait)
;;       (common:wait-for-cpuload maxload numcpus waitdelay count: (- count 1) msg: msg remote-host: remote-host))
;;      (else
;;       (if (> num-tries 0)
;; 	  (if (common:low-noise-print 30 (conc (round first) "-load-acceptable-" (or remote-host "localhost")))
;; 	      (debug:print 0 *default-log-port* "Load on " (or remote-host "localhost") " is acceptable at " first " continuing."))
;; 	  (debug:print 0 *default-log-port* "Load on " (or remote-host "localhost") ", "first" could not be retrieved. Giving up and continuing."))))))
;; 
(define (get-uname . params)
  (let* ((uname-res (process:cmd-run->list (conc "uname " (if (null? params) "-a" (car params)))))
	 (uname #f))
    (if (null? (car uname-res))
	"unknown"
	(caar uname-res))))

;; for reasons I don't understand multiple calls to real-path in parallel threads
;; must be protected by mutexes
;;
(define (common:real-path inpath)
  ;; (process:cmd-run-with-stderr->list "readlink" "-f" inpath)) ;; cmd . params)
  ;; (let-values 
  ;;  (((inp oup pid) (process "readlink" (list "-f" inpath))))
  ;;  (with-input-from-port inp
  ;;    (let loop ((inl (read-line))
  ;;       	(res #f))
  ;;      (print "inl=" inl)
  ;;      (if (eof-object? inl)
  ;;          (begin
  ;;            (close-input-port inp)
  ;;            (close-output-port oup)
  ;;            ;; (process-wait pid)
  ;;            res)
  ;;          (loop (read-line) inl))))))
  (with-input-from-pipe (conc "readlink -f " inpath) read-line))

;;======================================================================
;; D I S K   S P A C E 
;;======================================================================

(define (common:get-disk-space-used fpath)

   ((equal? status "WARN")    "orange")
   ((equal? status "KILLED")  "orange")
   ((equal? status "KILLREQ") "purple")
   ((equal? status "RUNNING") "blue")
   ((equal? status "ABORT")   "brown")
   (else "black")))

;; ;;======================================================================
;; ;; N A N O M S G   C L I E N T
;; ;;======================================================================
;; 
;; 
;; 
;; (define (common:send-dboard-main-changed)
;;   (let* ((dashboard-ips (mddb:get-dashboards)))
;;     (for-each
;;      (lambda (ipadr)
;;        (let* ((soc (common:open-nm-req (conc "tcp://" ipadr)))
;; 	      (msg (conc "main " *toppath*))
;; 	      (res (common:nm-send-receive-timeout soc msg)))
;; 	 (if (not res) ;; couldn't reach that dashboard - remove it from db
;; 	     (print "ERROR: couldn't reach dashboard " ipadr))
;; 	 res))
;;      dashboard-ips)))
;;     
;;     
;; ;;======================================================================
;; ;; D A S H B O A R D   D B 
;; ;;======================================================================
;; 
;; (define (mddb:open-db)
;;   (let* ((db (open-database (conc (get-environment-variable "HOME") "/.dashboard.db"))))
;;     (set-busy-handler! db (busy-timeout 10000))
;;     (for-each
;;      (lambda (qry)
;;        (exec (sql db qry)))
;;      (list 
;;       "CREATE TABLE IF NOT EXISTS vars       (id INTEGER PRIMARY KEY,key TEXT, val TEXT, CONSTRAINT varsconstraint UNIQUE (key));"
;;       "CREATE TABLE IF NOT EXISTS dashboards (
;;           id         INTEGER PRIMARY KEY,
;;           pid        INTEGER,
;;           username   TEXT,
;;           hostname   TEXT,
;;           ipaddr     TEXT,
;;           portnum    INTEGER,
;;           start_time TIMESTAMP DEFAULT (strftime('%s','now')),
;;              CONSTRAINT hostport UNIQUE (hostname,portnum)
;;         );"
;;       ))
;;     db))
;; 
;; ;; register a dashboard 
;; ;;
;; (define (mddb:register-dashboard port)
;;   (let* ((pid      (current-process-id))
;; 	 (hostname (get-host-name))
;; 	 (ipaddr   (server:get-best-guess-address hostname))
;; 	 (username (current-user-name)) ;; (car userinfo)))
;; 	 (db      (mddb:open-db)))
;;     (print "Register monitor, pid: " pid ", hostname: " hostname ", port: " port ", username: " username)
;;     (exec (sql db "INSERT OR REPLACE INTO dashboards (pid,username,hostname,ipaddr,portnum) VALUES (?,?,?,?,?);")
;; 	   pid username hostname ipaddr port)
;;     (close-database db)))
;; 
;; ;; unregister a monitor
;; ;;
;; (define (mddb:unregister-dashboard host port)
;;   (let* ((db      (mddb:open-db)))
;;     (print "Register unregister monitor, host:port=" host ":" port)
;;     (exec (sql db "DELETE FROM dashboards WHERE hostname=? AND portnum=?;") host port)
;;     (close-database db)))
;; 
;; ;; get registered dashboards
;; ;;
;; (define (mddb:get-dashboards)
;;   (let ((db (mddb:open-db)))
;;     (query fetch-column
;; 	   (sql db "SELECT ipaddr || ':' || portnum FROM dashboards;"))))
    
;;======================================================================
;;  T E S T   L A U N C H I N G   P E R   I T E M   W I T H   H O S T   T Y P E S
;;======================================================================
;; 
;; [hosts]
;; arm cubie01 cubie02