[NIFI-5000] ListHDFS doesn't list files from updated 'directory'

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 1.2.0, 1.3.0, 1.4.0, 1.5.0, 1.6.0
• Fix Version/s: 1.7.0
• Component/s: Extensions
• Flags: Important
• Assignee: Sivaprasanna Sethuraman
• Reporter: Sivaprasanna Sethuraman
• Created: 21/Mar/18 16:38
• Updated: 25/Apr/18 16:08
• Resolved: 25/Apr/18 16:04

Description

ListHDFS lists files and saves the latest listed files' modified time - latestTimestampListed and latestTimestampEmitted in the `StateMap`. It is overriding `onPropertyModified` to check if the `Directory` or the `File Filter` has been modified and if they are indeed modified, it will reset the statemap variables to `-1L` so as to list all the files from the updated `Directory` or according to the updated `File Filter`. However it is not working as intended.

Scenario:

1. Create two directories in HDFS
   1. > hdfs dfs -mkdir /test1
   2. > hdfs dfs -mkdir /test2
2. Write files to the above directories in the following order:
   1. > hdfs dfs -put sample.txt /test1/t1_1.txt
   2. > hdfs dfs -put sample.txt /test2/t2_1.txt
   3. > hdfs dfs -put sample.txt /test1/t1_2.txt
3. Configure ListHDFS and set Directory to /test1 and start the processor. It will produce two flowfiles: t1_1.txt and t1_2.txt
4. Stop the processor. Configure and set Directory to /test2. Ideally the state variables (listed and emitted timestamp) should be reset and they should list the file t2_1.txt but it is not.
5. Now put one more file to test2:
   1. > hdfs dfs -put sample.txt /test2/2_2.txt
6. This would have listed the file t2_2.txt. File t2_1.txt is missed

Little debugging helped me found that the `onPropertyModified` indeed works as intended but somewhere else the code still reads the last saved state i.e. the modified time of /test1/t1_2.txt

Issue Links

• links to: GitHub Pull Request #2576