flume-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jeff Lord <jl...@cloudera.com>
Subject Re: best way to make all hdfs records in one file under a folder?
Date Mon, 20 Jan 2014 21:46:13 GMT
If you don't intend to roll based on # of events than you will want to set
rollCount to 0.
MyAgent.sinks.HDFS.hdfs.rollCount = 0


On Mon, Jan 20, 2014 at 12:35 PM, Jimmy <jimmyjack@gmail.com> wrote:

> Seems like the only reason is "too many files" issue, correct?
>
> File Crusher executed regularly might be better option than trying to tune
> it in flume
>
> http://www.jointhegrid.com/hadoop_filecrush/index.jsp
>
>
>
> ---------- Forwarded message ----------
> From: Chen Wang <chen.apache.solr@gmail.com>
> Date: Mon, Jan 20, 2014 at 11:21 AM
> Subject: Re: best way to make all hdfs records in one file under a folder?
> To: user@flume.apache.org
>
>
> Chris,
> Its by every 6 minutes(thats why i set the roll time to be 60*5=300. the
> data size is around 15M. Thus I want them all in one file.
> Chen
>
>
> On Mon, Jan 20, 2014 at 10:57 AM, Christopher Shannon <
> cshannon108@gmail.com> wrote:
>
>> How is your data partitioned, by date?
>>
>>
>> On Monday, January 20, 2014, Chen Wang <chen.apache.solr@gmail.com>
>> wrote:
>>
>>> Guys,
>>> I have flume setup to flow partitioned data to hdfs, each partition has
>>> its own file folder. Is there a way to specify all the data under one
>>> partition to be in one file?
>>> I am currently using
>>> MyAgent.sinks.HDFS.hdfs.batchSize = 10000
>>> MyAgent.sinks.HDFS.hdfs.rollSize = 15000000
>>> MyAgent.sinks.HDFS.hdfs.rollCount = 10000
>>> MyAgent.sinks.HDFS.hdfs.rollInterval = 360
>>>
>>> to make the file roll on 15m data or after 6 minute.
>>>
>>> Is this the best way to achieve my goal?
>>> Thanks,
>>> Chen
>>>
>>>
>
>
>

Mime
View raw message