[NIFI-7049] SFTP processors shouldn't silently try to access known hosts file of the user

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 1.10.0
• Fix Version/s: 1.12.0, 1.11.3
• Component/s: Extensions
• Assignee: Dayakar M
• Reporter: Arpad Boda
• Created: 21/Jan/20 13:59
• Updated: 21/Feb/20 20:37
• Resolved: 04/Feb/20 06:06

Description

In case NiFi test are executed on a machine without knows_hosts file, it's going to fail:

[INFO] Running org.apache.nifi.processors.standard.TestGetSFTP
[ERROR] Tests run: 2, Failures: 2, Errors: 0, Skipped: 0, Time elapsed: 0.374 s <<< FAILURE! - in org.apache.nifi.processors.standard.TestGetSFTP
[ERROR] testGetSFTPFileBasicRead(org.apache.nifi.processors.standard.TestGetSFTP)  Time elapsed: 0.132 s  <<< FAILURE!
java.lang.AssertionError: expected:<4> but was:<0>
	at org.apache.nifi.processors.standard.TestGetSFTP.testGetSFTPFileBasicRead(TestGetSFTP.java:88)

[ERROR] testGetSFTPIgnoreDottedFiles(org.apache.nifi.processors.standard.TestGetSFTP)  Time elapsed: 0.013 s  <<< FAILURE!
java.lang.AssertionError: expected:<2> but was:<0>
	at org.apache.nifi.processors.standard.TestGetSFTP.testGetSFTPIgnoreDottedFiles(TestGetSFTP.java:110)

Just pasting my private message that summarised this error previously:

https://github.com/apache/nifi/blob/master/nifi-nar-bundles/nifi-standard-bundle/nifi-standard-processors/src/main/java/org/apache/nifi/processors/standard/util/SFTPTransfer.java#L556
So the problem is that host key file is not a mandatory, but in case it’s not provided, we call load on the 3rd party lib without arguments:
https://github.com/hierynomus/sshj/blob/master/src/main/java/net/schmizz/sshj/SSHClient.java#L621
Which tries to load keys from the default location, but this is far from what we state in our documentation:

Host Key File            If supplied, the given file will be used as the Host Key; otherwise, no use host key file will be used 

So there are multiple issues here:
-Even though the ssh connection fails, somewhere the IO exception is swallowed. Didn’t reproduce to check the logs, but I would expect exceptions to be thrown in the testcase and these being talkative about the error. My gut feeling says that we do the same in case the user specifies a host key file, but it’s somehow not accessible.
-Strict host check on/off might not be enough to cover all the scenarios as there are three: host 1# known and key matches, 2# host not known and we either trust or not, 3# host known, but there is a mismatch (probably man in the middle). I think this property should be improved at least in documentation point of view as currently only the code tells what do we do in 2#. Which depends on whether the file exists or not, so most probably something unintended.
-Either the documentation or the behaviour should be fixed to make them aligned
-The testcase should either use a predefined key or have host key checking completely off. According to what we see above, not sure about the latter being nicely supported.

Issue Links

• links to: GitHub Pull Request #4014