Home | About | Sematext search-lucene.com search-hadoop.com
 Search Hadoop and all its subprojects:

Switch to Threaded View
Flume >> mail # user >> flume 0.9.4 windows version out of memory


Copy link to this message
-
Re: flume 0.9.4 windows version out of memory
+1

Thats the easiest option on Windows, unfortunately.

- Alex

On Jan 28, 2013, at 7:49 AM, Jeong-shik Jang <[EMAIL PROTECTED]> wrote:

> How about editing registry using regedit?
>
> 1. locate memory option parameters in registry:
> HKEY_LOCAL_MACHINE > SOFTWARE > Apache Software Foundation > Procrun 2.0 > FlumeNode > Parameters > Java
>
> 2. Increase values in JvmMs and JvmMx
>
> 3. Stop/Start FlumeNode Service via Flume tray icon menu.
>
> -JS
>
> On 1/28/13 1:05 PM, 周梦想 wrote:
>> Thank you, Jeong-shik!
>> I also consider that the OPENNING status related to ack rolling. Because it happened on Windows, not on Linux although I closed the firewall of windows. But I don't know why, maybe there are some anti-virus software do some thing to stop ack.
>>
>> The flume node tray icon can't modify the parameter of java VM. It's read only.
>>
>> Best regards!
>> Andy
>>
>> 2013/1/25 Jeong-shik Jang <[EMAIL PROTECTED] <mailto:[EMAIL PROTECTED]>>
>>
>>    Hi Andy,
>>
>>    Seeing your log, out of memory error had occurred during opening
>>    stage of the logical node, and so was the last status of the
>>    logical node reported to master; no more status was reported to
>>    master due to the error since then, so it stays as OPENNING in master.
>>    Your error seems to be related to ack rolling because agentSink is
>>    agentE2ESink; agentDFO is not related to ack but still WAL for
>>    disk failover, though.
>>
>>    I remember you can change java memory option in flume procrun tray
>>    (flumenodew.exe) icon menu; one looks just like Apache web server
>>    tray icon.
>>    You need to restart(stop and start) service through the tray menu
>>    once you make the change.
>>
>>    -JS
>>
>>
>>    On 1/25/13 5:34 PM, 周梦想 wrote:
>>>    I found that if I change windows agent configure from
>>>    tail("c:\\test.txt") agentSink("hadoop48",35853) ;
>>>
>>>    to
>>>    tail("c:\\test.txt") agentDFOSink("hadoop48",35853) ;
>>>
>>>
>>>    then the agent changes status from OPENING to ACTIVE.
>>>    my windows agent is :192.168.20.81, while linux collector is :
>>>    192.168.10.48
>>>    another linux agent is 192.168.10.46. 10.46 have no problem of
>>>    agentSink.
>>>
>>>    Thanks!
>>>    Andy.
>>>
>>>    2013/1/25 周梦想 <[EMAIL PROTECTED] <mailto:[EMAIL PROTECTED]>>
>>>
>>>        when i start the flume node service,the agent is OPENING.
>>>        Hello,
>>>
>>>        I need to gather logs from windows 7, but the flume master
>>>        and collector is running on linux. I download 0.9.4 exe file
>>>        from flume official web.
>>>
>>>        When I start the node from windows, the status is OPENING.
>>>
>>>        zhouhh.TKOffice.local zhouhh.TKOffice.local
>>>        zhouhh.TKOffice.local OPENING Fri Jan 25 11:45:01 CST 2013 4
>>>
>>>
>>>        And the log reports OutOfMemoryError.
>>>
>>>        ===============>>>        D:\Program Files\Cloudera\Flume
>>>        0.9.4\log\flumenode-stdout.2013-01-25.log
>>>
>>>        2013-01-25 15:31:03 Commons Daemon procrun stdout
>>>        initializedListening for transport dt_socket at address: 8888
>>>        java.lang.OutOfMemoryError: Java heap space
>>>        Dumping heap to java_pid8100.hprof ...
>>>        Heap dump file created [13270531 bytes in 0.115 secs]
>>>
>>>        ==============>>>        D:\Program Files\Cloudera\Flume
>>>        0.9.4\log\flumenode-stderr.2013-01-25.log
>>>
>>>        2013-01-25 15:31:03 Commons Daemon procrun stderr initialized
>>>        13/01/25 15:31:03 INFO agent.FlumeNode: Flume 0.9.4
>>>        13/01/25 15:31:03 INFO agent.FlumeNode:  rev
>>>        8d0aa4be4ee50bb664cb0d1624e5634f46d7b62a
>>>        13/01/25 15:31:03 INFO agent.FlumeNode: Compiled  on  Mon Jun
>>>        20 14:46:53 PDT 2011
>>>        13/01/25 15:31:03 INFO agent.FlumeNode: System property
>>>        java.runtime.name <http://java.runtime.name>=Java(TM) SE
>>>        Runtime Environment
>>>        13/01/25 15:31:03 INFO agent.FlumeNode
Alexander Alten-Lorenz
http://mapredit.blogspot.com
German Hadoop LinkedIn Group: http://goo.gl/N8pCF