| ︙ | | |
23
24
25
26
27
28
29
30
31
32
33
34
35
36
|
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
|
+
+
+
|
(declare (uses debugprint))
(declare (uses stml2))
(declare (uses pkts))
(declare (uses processmod))
(declare (uses mtargs))
(declare (uses configfmod))
;; odd but it works?
(declare (uses itemsmod))
(module commonmod
*
(import scheme
chicken.base
chicken.condition
chicken.file
|
| ︙ | | |
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
|
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
|
-
+
+
|
mtver
debugprint
stml2
pkts
processmod
(prefix mtargs args:)
configfmod
|
| ︙ | | |
96
97
98
99
100
101
102
103
104
105
106
107
108
109
|
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
|
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
|
(define home (getenv "HOME"))
(define user (getenv "USER"))
;; Globals
;;
(define *server-loop-heart-beat* (current-seconds))
;; copied from egg call-with-environment-variables
;;
(define (call-with-environment-variables variables thunk)
;; @("Sets up environment variable via dynamic-wind which are taken down after thunk."
;; (variables "An alist of the form {{'((\"var\" . \"value\") ...)}}")
;; (thunk "The thunk to execute with a modified environment"))
(let ((pre-existing-variables
(map (lambda (var-value)
(let ((var (car var-value)))
(cons var (get-environment-variable var))))
variables)))
(dynamic-wind
(lambda () (void))
(lambda ()
;; (use posix)
(for-each (lambda (var-value)
(setenv (car var-value) (cdr var-value)))
variables)
(thunk))
(lambda ()
(for-each (lambda (var-value)
(let ((var (car var-value))
(value (cdr var-value)))
(if value
(setenv var value)
(unsetenv var))))
pre-existing-variables)))))
;; 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))
(let ((fmod-time (handle-exceptions
ext
|
| ︙ | | |
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
|
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
|
-
+
|
(define *user-hash-data* (make-hash-table))
(define *db-keys* #f)
(define *pkts-info* (make-hash-table)) ;; store stuff like the last parent here
(define *configinfo* #f) ;; raw results from setup, includes toppath and table from megatest.config
(define *runconfigdat* #f) ;; run configs data
(define *configdat* #f) ;; megatest.config data
;; (define *configdat* #f) ;; megatest.config data ==> moved to configfmod
(define *configstatus* #f) ;; status of data; 'fulldata : all processing done, #f : no data yet, 'partialdata : partial read done
(define *toppath* #f)
(define *already-seen-runconfig-info* #f)
(define *test-meta-updated* (make-hash-table))
(define *globalexitstatus* 0) ;; attempt to work around possible thread issues
(define *passnum* 0) ;; when running track calls to run-tests or similar
|
| ︙ | | |
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
|
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
|
+
+
+
+
+
+
+
+
+
+
|
(begin
;; (print "NO MATCH: " hed)
(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)))
(define (launch:is-test-alive host pid)
(if (and host pid (not (equal? host "n/a")))
(let* ((cmd (conc "ssh " host " pstree -A " pid))
(output (with-input-from-pipe cmd read-lines)))
(debug:print 2 *default-log-port* "Running " cmd " received " output)
(if (eq? (length output) 0)
#f
#t))
#t))
;; common:get-host-info was here
;; common:update-host-loads-table
;; common:get-least-loaded-host
;; common:wait-for-homehost-load
(define (common:get-num-cpus remote-host)
|
| ︙ | | |
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
|
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
2189
2190
2191
2192
2193
2194
2195
2196
2197
2198
2199
2200
2201
2202
2203
2204
2205
2206
2207
2208
2209
2210
2211
2212
2213
2214
2215
2216
2217
2218
2219
2220
2221
2222
2223
2224
|
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
|
(else
(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 "
normalized-effective-load " continuing."))
(debug:print 0 *default-log-port* "Load on " effective-host ", "
first" could not be retrieved. Giving up and continuing."))))))
;;======================================================================
;; server process management
;;======================================================================
;; no elegance here ...
;;
(define (tasks:kill-server hostname pid #!key (kill-switch ""))
(debug:print-info 0 *default-log-port* "Attempting to kill server process " pid " on host " hostname)
(setenv "TARGETHOST" hostname)
(let* ((logdir (if (directory-exists? "logs")
"logs/"
""))
(logfile (if logdir (conc "logs/server-"pid"-"hostname".log") #f))
(gzfile (if logfile (conc logfile ".gz"))))
(setenv "TARGETHOST_LOGF" (conc logdir "server-kills.log"))
(system (conc "nbfake kill "kill-switch" "pid))
(when logfile
(thread-sleep! 0.5)
(if (file-exists? gzfile) (delete-file gzfile))
(system (conc "gzip " logfile))
(unsetenv "TARGETHOST_LOGF")
(unsetenv "TARGETHOST"))))
(define (server:get-logs-list area-path)
(let* ((server-logs-cmd (conc "grep -iL exiting " area-path "/logs/server-*-*.log"))
(server-logs (string-split (string-chomp (with-input-from-pipe server-logs-cmd read-string)))))
server-logs))
;; 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-writable? 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. exn=" exn)))
(directory-exists? (conc areapath "/logs")))
'()))
;; Get the list of server logs that do not contain "exiting". Ignore logs for servers that have exited.
(let* ((server-logs (server:get-logs-list areapath))
(num-serv-logs (length server-logs)))
(if (or (null? server-logs) (= num-serv-logs 0))
(let ()
(debug:print 1 *default-log-port* "There are no servers running")
'()
)
(let loop ((hed (string-chomp (car server-logs)))
(tal (cdr server-logs))
(res '()))
(let* ((mod-time (handle-exceptions
exn
(begin
(debug:print 0 *default-log-port* "failed to get modification time on " hed ", exn=" 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)))) ;; any changes to number of elements in new-res will dirctly affect server:record->url,server:record->id,server:kill,server:get-num-alive which uses match let
(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 (string-chomp (car tal)) (cdr tal) new-res)))))))))
(define (server:get-num-alive srvlst)
(let ((num-alive 0))
(for-each
(lambda (server)
(handle-exceptions
exn
(begin
(debug:print-info 0 *default-log-port* "Unable to get server start-time and/or mod-time from " server ", exn=" exn))
(match-let (((mod-time host port start-time server-id 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
(or (not (configf:lookup *configdat* "server" "runtime")) ;; skip if not set
(< (- now start-time)
(+ (- (string->number (configf:lookup *configdat* "server" "runtime"))
180)
(pseudo-random-integer 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 (pseudo-random-integer len)))
(list-ref srvrs idx))
#f)))
(define (server:record->id servr)
(handle-exceptions
exn
(begin
(debug:print-info 0 *default-log-port* "Unable to get server id from " servr ", exn=" exn)
#f)
(match-let (((mod-time host port start-time server-id pid)
servr))
(if server-id
server-id
#f))))
(define (server:get-num-servers #!key (numservers 2))
(let ((ns (string->number
(or (configf:lookup *configdat* "server" "numservers") "notanumber"))))
(or ns numservers)))
;; given a path to a server log return: host port startseconds
;; any changes to number of elements returned by this fuction will dirctly affect server:record->url,server:record->id,server:kill,server:get-num-alive which uses match let
(define (server:logf-get-start-info logf)
(let ((server-rx (regexp "^SERVER STARTED: (\\S+):(\\d+) AT ([\\d\\.]+) server-id: (\\S+)")) ;; SERVER STARTED: host:port AT timesecs server id
(dbprep-rx (regexp "^SERVER: dbprep"))
(dbprep-found 0))
(handle-exceptions
exn
(begin
(debug:print-info 0 *default-log-port* "Unable to get server info from " logf ", exn=" exn)
(list #f #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 server-rx inl))
(dbprep (string-match dbprep-rx inl))
)
(if dbprep
(set! dbprep-found 1)
)
(if (not mlst)
(if (< lnum 500) ;; give up if more than 500 lines of server log read
(loop (read-line)(+ lnum 1))
(begin
(debug:print-info 0 *default-log-port* "Unable to get server info from first 500 lines of " logf )
(list #f #f #f #f)))
(let ((dat (cdr mlst)))
(list (car dat) ;; host
(string->number (cadr dat)) ;; port
(string->number (caddr dat))
(cadr (cddr dat))))))
(begin
(if dbprep-found
(begin
(debug:print-info 0 *default-log-port* "Server is in dbprep at " (current-seconds))
(thread-sleep! 25)
)
(debug:print-info 0 *default-log-port* "Unable to get server info from " logf " at " (current-seconds))
)
(list #f #f #f #f)))))))))
;;======================================================================
;; 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
|
| ︙ | | |
3102
3103
3104
3105
3106
3107
3108
3109
3110
|
3364
3365
3366
3367
3368
3369
3370
3371
3372
3373
3374
3375
3376
3377
3378
3379
3380
3381
3382
3383
3384
3385
3386
3387
3388
3389
3390
3391
3392
3393
3394
3395
3396
3397
3398
3399
3400
3401
3402
3403
3404
3405
3406
3407
3408
3409
3410
3411
3412
3413
3414
3415
3416
3417
3418
3419
3420
3421
3422
3423
3424
3425
3426
3427
3428
3429
3430
3431
3432
3433
3434
3435
3436
3437
3438
3439
3440
3441
3442
3443
3444
3445
3446
3447
3448
3449
3450
3451
3452
3453
3454
3455
3456
3457
3458
3459
3460
3461
3462
3463
3464
3465
3466
3467
3468
3469
3470
3471
3472
3473
3474
3475
3476
3477
3478
3479
3480
3481
3482
3483
3484
3485
3486
3487
3488
3489
3490
3491
3492
3493
3494
3495
3496
3497
3498
3499
3500
3501
3502
3503
3504
3505
3506
3507
3508
3509
3510
3511
3512
3513
3514
3515
3516
3517
3518
3519
3520
3521
3522
3523
3524
3525
3526
3527
3528
3529
3530
3531
3532
3533
3534
3535
3536
3537
3538
3539
3540
3541
|
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
-
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
|
;; (debug:print-info 0 *default-log-port* "common-telemetry-log closure failure")
;; )
;; (begin
;; (define *common:telemetry-log-state* 'closed)
;; (udp-close-socket *common:telemetry-log-socket*)
;; (set! *common:telemetry-log-socket* #f)))))
;;======================================================================
;; test patt stuff
;;======================================================================
;; make a query (fieldname like 'patt1' OR fieldname
(define (db:patt->like fieldname pattstr #!key (comparator " OR "))
(let ((patts (if (string? pattstr)
(string-split pattstr ",")
'("%"))))
(string-intersperse (map (lambda (patt)
(let ((wildtype (if (substring-index "%" patt) "LIKE" "GLOB")))
(conc fieldname " " wildtype " '" patt "'")))
(if (null? patts)
'("")
patts))
comparator)))
)
|