archiva-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Benjamin Knoth <kn...@mpdl.mpg.de>
Subject Re: 100 %cpu usage
Date Fri, 11 Feb 2011 13:37:40 GMT
Hi,

I restarted on 11:06.

On 11:56 i got this messages.

2011-02-11 11:56:05,101 [btpool0-6] INFO
org.codehaus.plexus.redback.struts2.in
terceptor.ForceAdminUserInterceptor  - Admin user found. No need to
configure ad
min user.
2011-02-11 11:56:05,112 [btpool0-6] INFO
com.opensymphony.xwork2.validator.Acti
onValidatorManagerFactory  - Detected AnnotationActionValidatorManager,
initiali
zing it...


On 11:55:52 archiva runs fine
 PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND
6765 root      16   0  798m 326m  11m S  0.0  8.0   0:28.37 java

On 11:58:52 it runs with full cpu usage
  PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND

 6765 root      16   0  800m 346m  11m S 91.5  8.5   3:13.04 java

After 12:01:52 it runs with 100% cpu usage. After restart it's all fine.
  PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND

 6765 root      16   0  800m 353m  11m S 99.5  8.6   6:12.21 java

Best regards
Benjamin


Am 11.02.2011 11:32, schrieb Benjamin Knoth:
> I set the crons in achiva to check only 1 times a day in the night
> (0:15, 1:15 and 2:15). I monitored it with snapshots of top.
> 
> At 05:03:49 i got this:
> 17453 root      18   0  873m 725m  11m S  1.9 17.7  28:37.38 java
> 
> At 05:06:49 i got this:
> 17453 root      18   0  873m 725m  11m S 52.2 17.7  30:11.36 java
> 
> At 05:09:49 i got this:
>   PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND
> 
> 17453 root      18   0  873m 725m  11m S 99.8 17.7  33:11.01 java
> 
> On 5 o'clock there wasn't run a cronjob.
> 
> Today same problem a lot of faster.
> 
> Restart at 08:47:34
> 
> At 09:49:52
> 
>   PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND
> 
>  5504 root      15   0  856m 353m  11m S 99.8  8.6   3:19.09 java
> 
> We haven't any prpblems with swap.
> Swap:  4194296k total,       68k used,  4194228k free;
> 
> What is maestro and what does it do. I can't find it in the standard
> repository of sles..
> 
> What should i do with the YourKit Java Profiler.
> I haven't use this before.
> 
> Benjamin
> 
> 
> Am 09.02.2011 06:36, schrieb Brett Porter:
>>
>> On 08/02/2011, at 11:57 PM, Benjamin Knoth wrote:
>>
>>> Am 08.02.2011 12:38, schrieb Brett Porter:
>>>>
>>>> On 08/02/2011, at 8:00 PM, Benjamin Knoth wrote:
>>>>
>>>>> Hi Brett,
>>>>> this is the output.
>>>>
>>>> Everything looks normal here - you're not running out of memory, and the
scans are quick.
>>>>
>>>> You said the peaks coincided with the database scan in 1.3.1 - is that also
the case in 1.3.3? Do they still last 12 minutes?
>>>
>>> No the last output in the log file which i send you was repeated 4 times
>>> with a difference of +- 2 seconds. But since 5 hours it runs with 100%
>>> cpu usage.
>>
>> Sorry, I'd like to clarify I'm understanding correctly. What I've seen is the hourly
scans completing in ~1-2 minutes (10 seconds for the repository scan, 50 seconds for the database
scan). The status panel you showed had nothing queued. However, you're still seeing continuous
100% usage of the application for over 5 hours?
>>
>> How many requests is the box receiving? It doesn't appear you've got any problematic
background processing, so it might be request load.
>>
>> Or is it possible that something else on the VM is constraining the resources it
has? Or perhaps forcing swap space to be used?
>>
>>>
>>> It's possible that i need 2 cores of cpu or should i change the memory
>>> usage in archiva more than 1 GB ram. This VM have 4 GB Ram and 1 Core of
>>> the CPU.
>>
>> 1 core should be fine. The status page showed only 144M of the RAM had been allocated
(and only 48M used) out of the 991M you set aside, so that's not the issue.
>>
>> The only other investigation I can think of is to do a thread dump (./bin/maestro
dump). If you are wanting to analyze it closely you could connect a profiler (like yourkit)
to the running instance as well.
>>
>> - Brett
>>
>> --
>> Brett Porter
>> brett@apache.org
>> http://brettporter.wordpress.com/
>> http://au.linkedin.com/in/brettporter
>>
>>
>>
>>
> 

-- 
Benjamin Knoth
Max Planck Digital Library (MPDL)
Systemadministration
Amalienstrasse 33
80799 Munich, Germany
http://www.mpdl.mpg.de

Mail: knoth@mpdl.mpg.de
Phone:  +49 89 38602 202
Fax:    +49-89-38602-280

Mime
View raw message