Netflix uses Atlas [1], the standalone JAR is only 8 KB [2] but it's
written in Scala, not sure what the overall footprint of the Scala
libraries and other dependencies are...

Regards,
Matt

[1] https://github.com/netflix/atlas/wiki
[2] https://mvnrepository.com/artifact/com.netflix.atlas/atlas-standalone/1.4.4

On Fri, Aug 3, 2018 at 10:03 AM Joe Witt <[email protected]> wrote:
>
> are there any good local/embedded style time series type DBs we can
> use for a non transient option?
>
> On Fri, Aug 3, 2018 at 9:54 AM, Mark Payne <[email protected]> wrote:
> > Hey Phil,
> >
> > There are two properties that control this:
> >
> > nifi.components.status.repository.buffer.size=1440
> > nifi.components.status.snapshot.frequency=1 min
> >
> > The snapshot.frequency tells NiFi how often to take a 'snapshot' of the 
> > stats on your canvas and add to the 'Status History'.
> > The buffer.size says how many of these to hold in memory.
> >
> > So the default is to take a snapshot every minute and keep 1,440 minutes' 
> > worth (which is 24 hours). If you want to expand this
> > to a week, you can do so by making the buffer size larger or by making the 
> > snapshot frequency less often. Keep in mind, though,
> > that there are a lot of stats being collected so if you have a lot of 
> > processors, the amount of heap/memory that gets used up can
> > grow pretty quickly. I'd recommend changing the snapshot frequency to "5 
> > mins" and the buffer size to 2016 to make 7 days' worth
> > of stats.
> >
> > You'll also want to keep in mind that these stats are all stored in memory 
> > so if you restart NiFi, you will lose the stats that you've gathered.
> >
> > Of note, I am currently working on NIFI-5466 [1], which should 
> > significantly reduce the amount of memory needed to store the stats, 
> > especially
> > for processors that are stopped.
> >
> > Hope this helps!
> >
> > Thanks
> > -Mark
> >
> > [1] https://issues.apache.org/jira/browse/NIFI-5466
> >
> > On Aug 2, 2018, at 7:59 PM, Phil H 
> > <[email protected]<mailto:[email protected]>> wrote:
> >
> > Hi there,
> >
> > Is there a way to plot graphs longer than 24 hours? We are interested in
> > seeing things that happen over the course of a week or so
> >
> > Thanks,
> > Phil
> >

Reply via email to