284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
|
(sqlite3:for-each-row
(lambda (count)
(set! res count))
mdb
"SELECT count(id) FROM monitors WHERE last_update < (strftime('%s','now') - 300) AND username=?;"
(car (user-information (current-user-id))))
res))
;;======================================================================
;; server stuff that operates on the server log files
;;======================================================================
(define (server:get-num-servers #!key (numservers 2))
(let ((ns (string->number
(or (configf:lookup *configdat* "server" "numservers") "notanumber"))))
(or ns numservers)))
;; 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 900)) ;; day-seconds))
(server:logf-get-start-info hed)
'())) ;; don't waste time processing server files not touched in the 15 minutes if there are more than ten servers to look at
(serv-rec (cons mod-time serv-dat))
(fmatch (string-match fname-rx hed))
(pid (if fmatch (string->number (list-ref fmatch 2)) #f))
(new-res (if (null? serv-dat)
res
(cons (append serv-rec (list pid)) res))))
(if (null? tal)
(if (and limit
(> (length new-res) limit))
new-res ;; (take new-res limit) <= need intelligent sorting before this will work
new-res)
(loop (car tal)(cdr tal) new-res)))))))))
;; 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))))))))
(define (server:get-num-alive srvlst)
(let ((num-alive 0))
(for-each
(lambda (server)
(match-let (((mod-time host port start-time pid)
server))
(let* ((uptime (- (current-seconds) mod-time))
(runtime (if start-time
(- mod-time start-time)
0)))
(if (< uptime 5)(set! num-alive (+ num-alive 1))))))
srvlst)
num-alive))
;; given a list of servers get a list of valid servers, i.e. at least
;; 10 seconds old, has started and is less than 1 hour old and is
;; active (i.e. mod-time < 10 seconds
;;
;; 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))
(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)))
(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)))
;;
#;(define (tasks:start-monitor db mdb)
(if (> (tasks:get-num-alive-monitors mdb) 2) ;; have two running, no need for more
(debug:print-info 1 *default-log-port* "Not starting monitor, already have more than two running")
(let* ((megatestdb (conc *toppath* "/megatest.db"))
(monitordbf (conc (db:dbfile-path #f) "/monitor.db"))
|
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
|
284
285
286
287
288
289
290
291
292
293
294
295
296
297
|
(sqlite3:for-each-row
(lambda (count)
(set! res count))
mdb
"SELECT count(id) FROM monitors WHERE last_update < (strftime('%s','now') - 300) AND username=?;"
(car (user-information (current-user-id))))
res))
;;
#;(define (tasks:start-monitor db mdb)
(if (> (tasks:get-num-alive-monitors mdb) 2) ;; have two running, no need for more
(debug:print-info 1 *default-log-port* "Not starting monitor, already have more than two running")
(let* ((megatestdb (conc *toppath* "/megatest.db"))
(monitordbf (conc (db:dbfile-path #f) "/monitor.db"))
|