Hey Chad,

What do you have for the value of the 
"nifi.provenance.repository.max.storage.size" property?
We will often see this if the value is very small (the default is 1 GB, which 
is very small) and the volume
of data is reasonably high.

The way that the repo works, it writes to one file for a while, until that file 
reaches 100 MB or up to 30 seconds,
by default (configured via "nifi.provenance.repository.rollover.size" and 
"nifi.provenance.repository.rollover.time").
At that point, it rolls over to writing to a new file and adds the 
now-completed file to a queue. A background thread
is then responsible for compressing that completed file.

What can happen, though, if the max storage space is small is that the data can 
actually be aged off from the repository
before that background task attempts to compress it. That can result in either 
a FileNotFoundException or an EOFException
when trying to read the TOC file (depending on the timing of when the age-off 
happens). It could potentially occur on the
.prov file, in addition to, or instead of the .toc file.

So generally, the solution is to increase the max storage size. It looks like 
you have 130 GB on each partition and 2 partitions per
node, so 260 GB total per node that you can use for provenance. So I would set 
the max storage size to something like "200 GB".
Since it is a soft limit and it may use more disk space than that temporarily 
before shrinking back down, you'll want to give it a little
bit of wiggle room.

Thanks
-Mark


On Feb 13, 2019, at 1:31 PM, Chad Woodhead 
<[email protected]<mailto:[email protected]>> wrote:

Hey Joe,

Yes 
nifi.provenance.repository.implementation=org.apache.nifi.provenance.WriteAheadProvenanceRepository

Disk space is fine as well. I have dedicated mounts for provenance (as well all 
the repos have their own dedicated mounts):

nifi.provenance.repository.dir.default=/data/disk5/nifi/provenance_repository
nifi.provenance.repository.directory.provenance2=/data/disk6/nifi/provenance_repository

Both of these mounts have plenty of space and are only 1% full and have never 
become close to being filled up.

<image.png>

-Chad

On Wed, Feb 13, 2019 at 1:06 PM Joe Witt 
<[email protected]<mailto:[email protected]>> wrote:
Chad,

In your conf/nifi.properties please see what the implementation is for your 
provenance repository. This specied on

nifi.provenance.repository.implementation=org.apache.nifi.provenance.WriteAheadProvenanceRepository

Is that what you have?

The above error I believe could occur if the location where provenance is being 
written runs out of disk space.  It is important to ensure that prov is sized 
and on a partition alone where this wont happen.  This is also true for the 
flow file repo.  The content repo is more resilient to this by design but still 
you want all three repo areas on their own partitions as per best practices.

Thanks
Joe

On Wed, Feb 13, 2019 at 1:03 PM Chad Woodhead 
<[email protected]<mailto:[email protected]>> wrote:
I use the org.apache.nifi.provenance.WriteAheadProvenanceRepository and I am 
seeing the following error in my logs a lot and I can't view any provenance 
data in the UI:

2019-02-13 12:57:44,637 ERROR [Compress Provenance Logs-1-thread-1] 
o.a.n.p.s.EventFileCompressor Failed to read TOC File 
/data/disk5/nifi/provenance_repository/toc/158994812.toc
java.io.EOFException: null
        at 
org.apache.nifi.provenance.toc.StandardTocReader.<init>(StandardTocReader.java:48)
        at 
org.apache.nifi.provenance.serialization.EventFileCompressor.run(EventFileCompressor.java:93)
        at 
java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:511)
        at java.util.concurrent.FutureTask.run(FutureTask.java:266)
        at 
java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)
        at 
java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)
        at java.lang.Thread.run(Thread.java:745)

Any ideas on what could be going on?

-Chad

Reply via email to