httpd-bugs mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From bugzi...@apache.org
Subject DO NOT REPLY [Bug 15418] New: - Load Spikes
Date Mon, 16 Dec 2002 21:28:53 GMT
DO NOT REPLY TO THIS EMAIL, BUT PLEASE POST YOUR BUG 
RELATED COMMENTS THROUGH THE WEB INTERFACE AVAILABLE AT
<http://nagoya.apache.org/bugzilla/show_bug.cgi?id=15418>.
ANY REPLY MADE TO THIS MESSAGE WILL NOT BE COLLECTED AND 
INSERTED IN THE BUG DATABASE.

http://nagoya.apache.org/bugzilla/show_bug.cgi?id=15418

Load Spikes

           Summary: Load Spikes
           Product: Apache httpd-1.3
           Version: 1.3.27
          Platform: PC
               URL: http://www.network54.com/spikes/
        OS/Version: Other
            Status: NEW
          Severity: Normal
          Priority: Other
         Component: core
        AssignedTo: bugs@httpd.apache.org
        ReportedBy: sroussey@network54.com


I have been researching this issue for some time...

Relevant software versions:
  Apache is 1.3.27 with SSL (2.8.12)
  PHP 4.2.3 (w/ Zend P.S. 3.0.0)

We experience load spikes throughout the day. It almost looks counter driven -- 
that is, after a certain number of requests it goes into a spike. So at night 
they are farther between than in the middle of the day when things are busy. It 
is not really counter driven, of course, just looks that way. Look at 
http://www.network54.com/spikes/spikes.html for a graph.

The vmstat (at http://www.network54.com/spikes/vmstat.txt and included at the 
end of the message) is intriguing. Right before a spike, idle time soars and 
the number of processes running is zero. Then the opposite, as it seems all 
processes get woken up and idle time goes to zero. It then takes a while to 
unwind.

Additional note: During a spike, the parent seems to try and wake everyone up 
and then starts spawning more children until it reaches its limit. All apache 
processes seem to be writing (could be 100,500,1000, whatever the 
HARD_SERVER_LIMIT was). The spikes still occur even if the MinServers is set to 
the max and the parent process does not fork(). It is just worse when it does 
the fork as well.












   procs                      memory    swap          io     system         cpu
 r  b  w   swpd   free   buff  cache  si  so    bi    bo   in    cs  us  sy  id
35  0  0  15960 129416  30932 208580   0   0     0     0 6908 29026  50  33  16
58  0  0  15960 129636  30932 208580   0   0     0     0 4687 32111  24  31  45
95  0  2  15960 130004  30940 208600   0   0    24    52 6520 15570  41  26  33
88  0  2  15960 129084  30940 208600   0   0     0     0 8664 19536  64  36   0
93  0  2  15960 128152  30940 208648   0   0     0     0 8690 19290  67  33   0
 4  0  0  15960 126680  30940 208652   0   0     0     0 8455 20292  66  31   3
32  0  0  15960 127264  30940 208672   0   0     0     0 3492 19996  16  21  63
 2  0  0  15960 127596  30944 208676   0   0     0    84 3733 12757  20  17  63
 1  0  0  15960 127712  30944 208680   0   0     0     0 2474  4759  12   8  79
 0  0  0  15960 127824  30944 208688   0   0     0     0 3028  1248  15   8  77
 0  0  0  15960 127660  30944 208720   0   0     0     0 7436  3884  35  19  46
 0  0  0  15960 126924  30948 208724   0   0     0    56 1361   394   4   3  93
 0  0  0  15960 122576  30948 208732   0   0     0     0 1306   402   6   4  90
54  0  3  15960 109112  30948 208740   0   0     0     0 6100  2842  35  17  48
327  0  2  15960  91432  30948 208744   0   0     0     0 5795  2404  41  20  39
299  1  3  15960  79884  30948 208764   0   0     0     0 9154  3847  73  27   0
277  0  2  15960  78028  30964 208768   0   0    16   244 8637  3911  67  33   0
285  0  2  15960  74708  30964 208828   0   0     0     0 10260  4731  68  32   
0
275  0  2  15960  71660  30984 208880   0   0    68     0 13746  6097  66  34   
0
279  0  2  15960  70708  30996 208932   0   0    72     0 9439  4312  64  36   0
283  0  2  15960  68624  31004 208956   0   0    24    64 9800  4874  60  39   1
281  0  2  15960  67808  31004 208964   0   0     0     0 9773  4489  63  37   0
232  0  2  15960  66776  31004 208972   0   0     0     0 10394  6636  58  41   
0
223  0  2  15960  63644  31012 209040   0   0    72     0 11785 10343  55  45   
0
241  0  2  15960  63372  31012 209148   0   0   104     0 9064 12047  63  37   0

---------------------------------------------------------------------
To unsubscribe, e-mail: bugs-unsubscribe@httpd.apache.org
For additional commands, e-mail: bugs-help@httpd.apache.org


Mime
View raw message