Megatest

Check-in [3321c2c522]
Login
Overview
Comment:lowered priority on some messages from get/write info. Fixed variable name.
Downloads: Tarball | ZIP archive | SQL archive
Timelines: family | ancestors | descendants | both | v1.65
Files: files | file ages | folders
SHA1: 3321c2c52278ba7f8d12aaea3cbc07f6a505fedc
User & Date: matt on 2018-02-04 23:10:59
Other Links: branch diff | manifest | tags
Context
2018-02-05
09:25
Increased allowed age on cached load info to 5 seconds. Increased random server start delay to 0-2 seconds. check-in: 97727e82b8 user: mrwellan tags: v1.65
2018-02-04
23:10
lowered priority on some messages from get/write info. Fixed variable name. check-in: 3321c2c522 user: matt tags: v1.65
21:13
Use the hostname for host load cache when running get load locally. check-in: 067b4c3661 user: mrwellan tags: v1.65
Changes

Modified common.scm from [d0d48bccf9] to [3258e08dfd].

1435
1436
1437
1438
1439
1440
1441
1442

1443
1444
1445
1446
1447

1448
1449
1450

1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462


1463
1464
1465
1466
1467
1468
1469
1470










1471
1472
1473
1474
1475
1476

1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517









































1518
1519
1520


1521
1522
1523
1524
1525
1526
1527
1528
1529
1530

1531
1532
1533
1534
1535
1536
1537
1435
1436
1437
1438
1439
1440
1441

1442
1443
1444
1445
1446

1447
1448
1449

1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464








1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481









































1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522



1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533

1534
1535
1536
1537
1538
1539
1540
1541







-
+




-
+


-
+












+
+
-
-
-
-
-
-
-
-
+
+
+
+
+
+
+
+
+
+






+
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
-
-
-
+
+









-
+







(define (common:get-cached-info key dtype #!key (age 3))
  (let* ((fullpath (conc *toppath* "/logs/" key "-" dtype ".log")))
    (if (and (file-exists? fullpath)
	     (file-read-access? fullpath))
	(handle-exceptions
	 exn
	 #f
	 (print "reading file " fullpath)
	 (debug:print 2 *default-log-port* "reading file " fullpath)
	 (let ((real-age (- (current-seconds)(file-change-time fullpath)))) 
	   (if (< real-age age)
	       (with-input-from-file fullpath read)
	       (begin
		 (print "file " fullpath " is too old (" real-age" seconds)to trust, skipping reading it")
		 (debug:print 2 *default-log-port* "file " fullpath " is too old (" real-age" seconds)to trust, skipping reading it")
		 #f))))
	(begin
	  (print "not reading file " fullpath)
	  (debug:print 2 *default-log-port* "not reading file " fullpath)
	  #f))))
 
(define (common:write-cached-info key dtype dat)
  (let* ((fullpath (conc *toppath* "/logs/" key "-" dtype ".log")))
    (handle-exceptions
     exn
     #f
     (with-output-to-file fullpath (lambda ()(pp dat))))))

;; get cpu load by reading from /proc/loadavg, return all three values
;;
(define (common:get-cpu-load remote-host)
  (let* ((actual-hostname (or remote-host (get-host-name))))
    (or (common:get-cached-info actual-hostname "cpu-load")
  (if remote-host
      (map (lambda (res)
	     (if (eof-object? res) 9e99 res))
	   (with-input-from-pipe 
	    (conc "ssh " remote-host " cat /proc/loadavg")
	    (lambda ()(list (read)(read)(read)))))
      (with-input-from-file "/proc/loadavg" 
	(lambda ()(list (read)(read)(read))))))
	(let ((result (if remote-host
			  (map (lambda (res)
				 (if (eof-object? res) 9e99 res))
			       (with-input-from-pipe 
				   (conc "ssh " remote-host " cat /proc/loadavg")
				 (lambda ()(list (read)(read)(read)))))
			  (with-input-from-file "/proc/loadavg" 
			    (lambda ()(list (read)(read)(read)))))))
	  (common:write-cached-info actual-hostname "cpu-load" result)
	  result))))

;; get normalized cpu load by reading from /proc/loadavg and /proc/cpuinfo return all three values and the number of real cpus and the number of threads
;; returns alist '((adj-cpu-load . normalized-proc-load) ... etc.
;;  keys: adj-proc-load, adj-core-load, 1m-load, 5m-load, 15m-load
;;
(define (common:get-normalized-cpu-load remote-host)
  (let ((actual-host (or remote-host (get-host-name)))) ;; #f is localhost
  (or (and remote-host (common:get-cached-info remote-host "normalized-load"))
      (let ((data (if remote-host
		      (with-input-from-pipe 
		       (conc "ssh " remote-host " cat /proc/loadavg;cat /proc/cpuinfo;echo end")
		       read-lines)
		      (append 
		       (with-input-from-file "/proc/loadavg" 
			 read-lines)
		       (with-input-from-file "/proc/cpuinfo"
			 read-lines)
		       (list "end"))))
	    (load-rx  (regexp "^([\\d\\.]+)\\s+([\\d\\.]+)\\s+([\\d\\.]+)\\s+.*$"))
	    (proc-rx  (regexp "^processor\\s+:\\s+(\\d+)\\s*$"))
	    (core-rx  (regexp "^core id\\s+:\\s+(\\d+)\\s*$"))
	    (phys-rx  (regexp "^physical id\\s+:\\s+(\\d+)\\s*$"))
	    (max-num  (lambda (p n)(max (string->number p) n))))
	;; (print "data=" data)
	(if (null? data) ;; something went wrong
	    #f
	    (let loop ((hed      (car data))
		       (tal      (cdr data))
		       (loads    #f)
		       (proc-num 0)  ;; processor includes threads
		       (phys-num 0)  ;; physical chip on motherboard
		       (core-num 0)) ;; core
	      ;; (print hed ", " loads ", " proc-num ", " phys-num ", " core-num)
	      (if (null? tal) ;; have all our data, calculate normalized load and return result
		  (let* ((act-proc (+ proc-num 1))
			 (act-phys (+ phys-num 1))
			 (act-core (+ core-num 1))
			 (adj-proc-load (/ (car loads) act-proc))
			 (adj-core-load (/ (car loads) act-core))
			 (result
			  (append (list (cons 'adj-proc-load adj-proc-load)
					(cons 'adj-core-load adj-core-load))
				  (list (cons '1m-load (car loads))
					(cons '5m-load (cadr loads))
					(cons '15m-load (caddr loads)))
				  (list (cons 'proc act-proc)
					(cons 'core act-core)
					(cons 'phys act-phys))))
    (or (common:get-cached-info actual-host "normalized-load")
	(let ((data (if remote-host
			(with-input-from-pipe 
			    (conc "ssh " remote-host " cat /proc/loadavg;cat /proc/cpuinfo;echo end")
			  read-lines)
			(append 
			 (with-input-from-file "/proc/loadavg" 
			   read-lines)
			 (with-input-from-file "/proc/cpuinfo"
			   read-lines)
			 (list "end"))))
	      (load-rx  (regexp "^([\\d\\.]+)\\s+([\\d\\.]+)\\s+([\\d\\.]+)\\s+.*$"))
	      (proc-rx  (regexp "^processor\\s+:\\s+(\\d+)\\s*$"))
	      (core-rx  (regexp "^core id\\s+:\\s+(\\d+)\\s*$"))
	      (phys-rx  (regexp "^physical id\\s+:\\s+(\\d+)\\s*$"))
	      (max-num  (lambda (p n)(max (string->number p) n))))
	  ;; (print "data=" data)
	  (if (null? data) ;; something went wrong
	      #f
	      (let loop ((hed      (car data))
			 (tal      (cdr data))
			 (loads    #f)
			 (proc-num 0)  ;; processor includes threads
			 (phys-num 0)  ;; physical chip on motherboard
			 (core-num 0)) ;; core
		;; (print hed ", " loads ", " proc-num ", " phys-num ", " core-num)
		(if (null? tal) ;; have all our data, calculate normalized load and return result
		    (let* ((act-proc (+ proc-num 1))
			   (act-phys (+ phys-num 1))
			   (act-core (+ core-num 1))
			   (adj-proc-load (/ (car loads) act-proc))
			   (adj-core-load (/ (car loads) act-core))
			   (result
			    (append (list (cons 'adj-proc-load adj-proc-load)
					  (cons 'adj-core-load adj-core-load))
				    (list (cons '1m-load (car loads))
					  (cons '5m-load (cadr loads))
					  (cons '15m-load (caddr loads)))
				    (list (cons 'proc act-proc)
					  (cons 'core act-core)
					  (cons 'phys act-phys)))))
                         (the-host (or remote-host (current-host-name))))
		    (common:write-cached-info the-host "normalized-load" result)
		    result)
		      (common:write-cached-info actual-host "normalized-load" result)
		      result)
		  (regex-case
		   hed
		   (load-rx  ( x l1 l5 l15 ) (loop (car tal)(cdr tal)(map string->number (list l1 l5 l15)) proc-num phys-num core-num))
		   (proc-rx  ( x p         ) (loop (car tal)(cdr tal) loads           (max-num p proc-num) phys-num core-num))
		   (phys-rx  ( x p         ) (loop (car tal)(cdr tal) loads           proc-num (max-num p phys-num) core-num))
		   (core-rx  ( x c         ) (loop (car tal)(cdr tal) loads           proc-num phys-num (max-num c core-num)))
		   (else 
		    (begin
		      ;; (print "NO MATCH: " hed)
		      (loop (car tal)(cdr tal) loads proc-num phys-num core-num))))))))))
		      (loop (car tal)(cdr tal) loads proc-num phys-num core-num)))))))))))

(define (common:unix-ping hostname)
  (let ((res (system (conc "ping -c 1 " hostname " > /dev/null"))))
    (eq? res 0)))

;; ideally put all this info into the db, no need to preserve it across moving homehost
;;
1640
1641
1642
1643
1644
1645
1646


1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660


















1661
1662
1663
1664
1665
1666
1667
1644
1645
1646
1647
1648
1649
1650
1651
1652














1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677







+
+
-
-
-
-
-
-
-
-
-
-
-
-
-
-
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+







                     #f
                     (common:get-homehost)))
         (hh     (if hh-dat (car hh-dat) #f))
         (numcpus (common:get-num-cpus hh)))
    (common:wait-for-normalized-load maxload msg hh)))

(define (common:get-num-cpus remote-host)
  (let* ((actual-host (or remote-host (get-host-name))))
    (or (common:get-cached-info actual-host "num-cpus" age: 86400) ;; hosts had better not be changing the number of cpus too often!
  (let ((proc (lambda ()
		(let loop ((numcpu 0)
			   (inl    (read-line)))
		  (if (eof-object? inl)
		      numcpu
		      (loop (if (string-match "^processor\\s+:\\s+\\d+$" inl)
				(+ numcpu 1)
				numcpu)
			    (read-line)))))))
    (if remote-host
	(with-input-from-pipe 
	 (conc "ssh " remote-host " cat /proc/cpuinfo")
	 proc)
	(with-input-from-file "/proc/cpuinfo" proc))))
	(let* ((proc   (lambda ()
			 (let loop ((numcpu 0)
				    (inl    (read-line)))
			   (if (eof-object? inl)
			       (begin
				 (common:write-cached-info remote-host "num-cpus" numcpu)
				 numcpu)
			       (loop (if (string-match "^processor\\s+:\\s+\\d+$" inl)
					 (+ numcpu 1)
					 numcpu)
				     (read-line))))))
	       (result (if remote-host
			   (with-input-from-pipe 
			       (conc "ssh " remote-host " cat /proc/cpuinfo")
			     proc)
			   (with-input-from-file "/proc/cpuinfo" proc))))
	  (common:write-cached-info actual-host "num-cpus" result)
	  result))))

;; wait for normalized cpu load to drop below maxload
;;
(define (common:wait-for-normalized-load maxload msg remote-host)
  (let ((num-cpus (common:get-num-cpus remote-host)))
    (common:wait-for-cpuload maxload num-cpus 15 msg: msg remote-host: remote-host)))

Modified tests/fdktestqa/testqa/runconfigs.config from [346ed47154] to [69a81e5b8c].



1
2
3
4
5
6
1
2
3
4
5
6
7
8
+
+






[include local.runconfigs]

[default]
ALLTESTS see this variable

# Your variables here are grouped by targets [SYSTEM/RELEASE]
[SYSTEM_val/RELEASE_val]
ANOTHERVAR only defined if target is SYSTEM_val/RELEASE_val