[
https://issues.apache.org/jira/browse/NIFI-3671?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15957006#comment-15957006
]
Joe Skora commented on NIFI-3671:
---------------------------------
[~markap14]
* It's an instance I built for testing last week based on master as of 3/26,
commit a266893 [1].
* The exception seems to keep the processors from starting, so I have to click
start again.
* What I say the graph is empty I mean that there are no files in any queues
anywhere in the graph.
* The system is an AWS "m4.2xlarge"[2] VM instance with 8 vCPU, 32GB, 500GB on
3 SSD volumes. There are 3 GenerateFlowFile processors creating 327680,
163840, and 163840 files of 1B each on a 24 hour Run Schedule so they only run
once. If I cut the Batch Size 1/4th and run 4 threads some threads might start
but rarely all 4 and then it's hard to get the desired number. Once it starts
to happen I can cut the Batch Size to 1/10 and it still will occur.
I'll try to get a template up shortly.
[1]https://git-wip-us.apache.org/repos/asf?p=nifi.git;a=commit;h=a26689318d68441f1e55b62e8ba64aab5bdfe053
[2]https://aws.amazon.com/ec2/instance-types/
> GenerateFlowFile throws NPE in FileSystemRepository
> ---------------------------------------------------
>
> Key: NIFI-3671
> URL: https://issues.apache.org/jira/browse/NIFI-3671
> Project: Apache NiFi
> Issue Type: Bug
> Components: Core Framework
> Affects Versions: 1.2.0
> Environment: Ubuntu Linux 14.04.5 LTS, x86, NiFi 1.2.0-SNAPSHOT, Java
> 1.8.0_102
> Reporter: Joe Skora
> Priority: Critical
> Fix For: 1.2.0
>
> Attachments: genff-npe.log
>
>
> While volume testing, {{GenerateFlowFile}} is randomly throwing
> {{NullPointerException}} during startup. The NPE occurs in
> {{FileSystemRepository.write()}} called from
> {{BufferedOutputStream.flushBuffer()}} and results in an Administrative Yield.
> When it happens, the bulletin shows up on the processors immediately after
> starting them. Starting them a second time will often succeed, but sometimes
> it takes a multiple retries. When the {{GenerateFlowFile}} processors are
> started for these tests, the graph is typically empty.
> Log entries for an example Thread are attached. This looks similar to
> [NIFI-2551|https://issues.apache.org/jira/browse/NIFI-2551], but that is
> marked as being fixed in 1.0.0.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
