Oleg, Yes, you are on the right track. The question is how to use custom classes on existing NiFi bundles / release packages without creating a code mess.
>From what I understand I could simply use Maven to compile a MapR friendly version of NiFi but some solutions are happy with basic mods to existing code: https://www.balabit.com/sites/default/files/documents/syslog-ng-ose-latest-guides/en/syslog-ng-ose-guide-admin/html/destination-hdfs-maprfs.html https://www-304.ibm.com/support/knowledgecenter/SSGSMK_7.1.0/mapreduce_integration/mapr.dita http://answers.mapr.com/questions/163588/reading-maprfs-from-spark.html Is my question a bit clearer now? Cheers On Tue, Oct 13, 2015 at 12:46 AM, Oleg Zhurakousky <[email protected]> wrote: > May be I am missing certain details, but are you asking about custom bundle > (NAR) or using existing NiFi bundles which would somehow have to delegate to > some custom JARs? > I am thinking that if you are using custom client, you may need a custom > bundle. > > Oleg > >> On Oct 12, 2015, at 9:29 AM, Andre F de Miranda <[email protected]> wrote: >> >> Hi there, >> >> First post in here so please... you know the drill. >> >> I was wondering: What is the recommended approach to add custom >> classes to NiFi on load time? >> >> Reason I ask is simple: Our HDFS store is in fact a MapR-FS store and >> as such I need to use the proprietary MapR client and Hadoop JARs on >> the client software. >> >> Usually this is achieved via add-on shims, rebuild (using maven >> artifacts) and in some rare cases, fully supported out of the box! :-) >> >> I am finding a bit difficult to setup the NiFI <-> MapR linkage >> without tainting the NiFi distribution folder structure and was >> wondering what is the recommended approach? >> >> I thank you in advance >> >
