And I'm back.

So, I had my processor running, no errors, so things were good.

I then add 2 if statements with logger messages, run the same mvn command, 
copied the new NAR on top of the old one, same versionId and all, and now I get 
a class not found error, that didn't occur before I stopped, redeployed and 
started. And I can't delete the event in the queue even though source and 
destination are stopped.

Is there problems when redeploying a custom processor NAR on top of itself? 
Even if the NiFi instance is stopped? Serialization issues?

Frustratedly yours,
John McGinn

--------------------------------------------
On Tue, 10/16/18, John McGinn <figgie...@sbcglobal.net> wrote:

 Subject: Re: Compiling custom processor
 To: dev@nifi.apache.org
 Date: Tuesday, October 16, 2018, 1:21 PM
 
 Thanks Bryan,
 
 I restarted with a fresh maven
 generate, and only added the API to the processor pom, and
 then the API-NAR to the NAR pom. I then figured out a
 subsequent dependency (lang3), and got that in, and now my
 processor shows up in local NiFi instance.
 
 Regarding the copying of the standard
 processor, when I copied Wait over, I renamed it from Wait
 to MyProcessorWait to be unique, and not copied. I'm just
 attempting changes to Wait without modifying the actual one.
 Also, thanks on the META-INF/services file information.
 
 Now I can debug and test.
 
 Sincerely,
 John McGinn
 
 --------------------------------------------
 On Tue, 10/16/18, Bryan Bende <bbe...@gmail.com>
 wrote:
 
  Subject: Re: Compiling custom
 processor
  To: dev@nifi.apache.org
  Date: Tuesday, October 16, 2018, 9:52
 AM
  
  In general, if your processor
  uses a controller service then the
  processors pom file needs a provided
 dependency
  on the API of the CS,
  and your NAR pom needs
  a NAR dependency on the NAR where the
 CS API
  is.
  
  Example is
  shown here in the section linking
 processors and
  controller
  services - 
https://cwiki.apache.org/confluence/display/NIFI/Maven+Projects+for+Extensions
  
  Also, you do not want to
  include NiFI's standard processors in
 your
  own NAR, this will result in two
 copies of
  every standard processor.
  You will want to
  copy whatever code you need into your
 own NAR.
  
  In you last statement, if you
  made a new processor in
  nifi-standard-processors and it didn't
 show
  up, it is most likely
  because you didn't
  update META-INF/services file to
 include the new
  processor.
  On Tue, Oct 16, 2018 at
  9:36 AM John McGinn
  <amruginn-n...@yahoo.com.invalid>
  wrote:
  >
  > Ok, I am
  far from a maven expert, and am
 struggling on this
  problem.
  >
  > I created
  a new project using the maven generate
 process, and compiled
  that sample processor and everything
 was fine, and I could
  see it in my local NiFi instance. I
 then copied over the
  Wait processor on top of the
 MyProcessor.java class, changed
  the package name, and the class name,
 and attempted to
  compile. Got errors due to dependency
 issues. This is where
  I get confused.
  >
  > In
  the processor directory, I modify the
 pom.xml with a
  dependency line for nifi-api,
 nifi-utils,
 
 nifi-distribute-cache-client-service-api and
  nifi-standard-processors. This seemed
 to make things work,
  and I get a 33 meg NAR file, which
 contains lots of JARs. I
  load that up, and I get a problem with
 CalculateStats, or
  similar, because of
 RecordReaderFactory. I go back in and
  include
 nifi-record-serialization-services, and it compiles,
  and the JAR is included in the NAR
 file now, and I still get
  class not found.
  >
  >
  I've also tried to change the
 dependencies to a status
  of provided, and therefore get a
 minimal 25k or so NAR file
  with no JARs included. I get the same
 issue with the class
  not found.
  >
  > I had
  also attempted to copy Wait.java to
 AnotherWait.java within
  the actually NiFi source code, and it
 compiles, and shows up
  in the JARs and NARs as I'd expect,
 but I cannot get to
  it within my NiFi instance.
  >
  > Can someone let me know what I'm
 doing
  wrong.
  >
  > Thanks,
  > John McGinn
 

Reply via email to