[jira] [Created] (NIFI-2199) NiFi cannot be restarted through SSH (nifi.sh hangs)

["Michal Klempa (JIRA)" <ji...@apache.org>]

Michal Klempa created NIFI-2199:
-----------------------------------

             Summary: NiFi cannot be restarted through SSH (nifi.sh hangs)
                 Key: NIFI-2199
                 URL: https://issues.apache.org/jira/browse/NIFI-2199
             Project: Apache NiFi
          Issue Type: Bug
    Affects Versions: 0.6.1, 0.5.1
            Reporter: Michal Klempa
            Priority: Minor


When NiFi is to restarted using SSH from remote host, nifi.sh causes SSH to hang on open io streams.

Steps to reproduce:
1. HostA - NiFi instalation
2. HostB - client machine
3. log to HostB
4. issue a command
{code}
ssh root@HostA /etc/init.d/nifi restart
{code}

Expected behavior: NiFi is restarted, SSH command exits with 0
Actual behavior: NiFi is restarted, SSH command hang on

This seems to be somehow related to SSH handling IO streams of process. Seems like detached NiFi java process does not close stdin/out/err (i do not know precisely) leading to sshd waiting for some input/output to come to client.

I was able to do a quick&dirty fix in nifi.sh as follows:
1. on existing nifi installation, in file /etc/init.d/nifi find the line
{code}
(cd "${NIFI_HOME}" && ${sudo_cmd_prefix} "${JAVA}" -cp "${BOOTSTRAP_CLASSPATH}" -Xms12m -Xmx24m -Dorg.apache.nifi.bootstrap.config.file="${BOOTSTRAP_CONF}" org.apache.nifi.bootstrap.RunNiFi $@ &)
{code}
its line number 214 (true branch of if [ "$1"] = "start"] condition).
2. replace it by line:
{code}
(cd "${NIFI_HOME}" && ${sudo_cmd_prefix} "${JAVA}" -cp "${BOOTSTRAP_CLASSPATH}" -Xms12m -Xmx24m -Dorg.apache.nifi.bootstrap.config.file="${BOOTSTRAP_CONF}" org.apache.nifi.bootstrap.RunNiFi $@ > /dev/null 2>&1 < /dev/null &)
{code}
This simply redirects stdin from /dev/null to NiFi java process and redirects stdout/err to /dev/null letting SSH to close the session.

After applying this fix, I am able to restart NiFi remotely using SSH.

Use case for this scenario: When Flow Manager develop flow.xml.gz on localhost/testing environments, usually deployment to production environment is handled by jenkins/bamboo CI tools. These tools need to log in, upload new flow.xml.gz and restart NiFi. All remotely using SSH.

If somebody can advice me where in the source code is this line of code located, I can prepare a pull request for this quick&dirty fix.
But. Maybe we may want to go further and investigate what to do with IO streams inside Java...





--
This message was sent by Atlassian JIRA
(v6.3.4#6332)