You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@flume.apache.org by Abhishek Pathak <ab...@gmail.com> on 2011/09/14 15:03:08 UTC
Can logical nodes have custom configurations?
Hi,
Is it possible for two logical nodes on the same physical node to support
different configurations?
I am specifically interested in creating a parallel flow in addition to the
existing flow.The existing flow rolls a new file every hour in the
HDFS,tailing the log file.I would like the new flow to tail the same
file,but roll a new file every minute in the local filesystem.I intend to
use this flow for real-time monitoring and error prediction in logs.
If there is a better way to achieve the same result,i am very interested.
Thanks,
Abhishek
Re: Can logical nodes have custom configurations?
Posted by Eric Sammer <es...@cloudera.com>.
Abishek:
You can definitely do this one of two ways. Syntactically, the way you get
two logical nodes on a single physical node is as follows.
ln1 : tail(...) | someSink(...);
ln2 : tail(...) | someOtherSink(...);
map physicalNode ln1;
map physicalNode ln2;
You see that ln1 and 2 can have different (or the same) sources and / or
sinks.
The other, more efficient, way of getting the same result is to use a fan
out sink. This splits the data stream in two copies.
ln1 : tail(...) | [ sink1(...), sink2(...) ] ;
This sends the same data to both sinks 1 and 2 without doing twice the tail
work. Check out the user guide for exact syntax examples, but you get the
idea.
On Wed, Sep 14, 2011 at 6:03 AM, Abhishek Pathak <
abhishek.pathak.iitk@gmail.com> wrote:
> Hi,
>
> Is it possible for two logical nodes on the same physical node to support
> different configurations?
> I am specifically interested in creating a parallel flow in addition to the
> existing flow.The existing flow rolls a new file every hour in the
> HDFS,tailing the log file.I would like the new flow to tail the same
> file,but roll a new file every minute in the local filesystem.I intend to
> use this flow for real-time monitoring and error prediction in logs.
> If there is a better way to achieve the same result,i am very interested.
>
> Thanks,
> Abhishek
>
--
Eric Sammer
twitter: esammer
data: www.cloudera.com