flume-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jonathan Hsieh <...@cloudera.com>
Subject Re: diskfailover
Date Mon, 01 Aug 2011 17:20:22 GMT
Hi Torsten,

Ideally, you would be using a agentDFOSink instead of the diskFailover.
  diskFailover decorator is used used internally with in it.  I've created a
page on the new wiki that talks about feature status and the intent of
certain sources and sinks.

Here's the feature status page:


On Wed, Jul 27, 2011 at 1:22 PM, Torsten Curdt <tcurdt@vafer.org> wrote:

> I am not sure I understand the semantics of the diskfailover decorator
> correctly.
>  { decorator(...) => { diskFailover(10000) => sink(...) }};
> If the sink throws an exception, the diskFailover should store the
> events to disk.
> That seems to work fine. But what about retry and replaying the events
> from disk?
> I thought I saw that working in 0.9.1 ...not anymore in 0.9.4
> Looking at the source I found that the diskFailover decorator supports
> two arguments, maxMillis and checkMillis. Unfortunately it's also not
> quite obvious what that means exactly.
> I assume/hope one of them is the delay between retries?
> Is that somewhere in the docs and I didn't find it?
> cheers,
> Torsten

// Jonathan Hsieh (shay)
// Software Engineer, Cloudera
// jon@cloudera.com

View raw message