[jira] [Commented] (NIFI-9511) sFTP connection resets

["David Handermann (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/NIFI-9511?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17463984#comment-17463984 ] 

David Handermann commented on NIFI-9511:
----------------------------------------

Thanks for reporting this issue and including the logs [~langheld].  Reviewing {{{}sshd-log.log{}}}, there are appear to be 61 connection attempts:


{noformat}
Connection from 10.32.28.37 port 59834 on 10.32.2.25 port 22 {noformat}

That is a lot of connections in a very short period of time.  Can you provide the number of Concurrent Tasks configured for the various SFTP processors?  There will be one SSH connection per Concurrent Task for each processor, so multiple Concurrent Tasks also increased the number of simultaneous connections.  Have you considered options for refactoring your flow to reduce the number of SFTP processors calling the same sever?  

There have been a number of changes to the SFTP processors since NiFi 1.11.4, including a change of the underlying SSH library.  It sounds like something may have changed related to connection reuse, which will involve further investigation.

> sFTP connection resets
> ----------------------
>
>                 Key: NIFI-9511
>                 URL: https://issues.apache.org/jira/browse/NIFI-9511
>             Project: Apache NiFi
>          Issue Type: Bug
>          Components: Core Framework
>    Affects Versions: 1.15.1
>            Reporter: Christoph Langheld
>            Priority: Major
>         Attachments: nifi-app.log, sshd-log.log
>
>
> Hello,
> with NiFi 1.15.1 we have many connection resets with processors GetSFTP and PutSFTP.
> The problem only seems to occur, if all the GetSFTP processors are running (about 20 that poll the same sever).
> Within sshd_config on the remote server we set MaxSessions to 50 and MaxStartups to 50;30;100.
> But the problem still occurs.
> With NiFi 1.11.4 we did not have that problem, so we think it's not  a problem of the remote server.



--
This message was sent by Atlassian Jira
(v8.20.1#820001)