CentOS 6.10 with Python 2.6.6 fail at reboot (overwriting env vars) about convert2rhel

[GeorgeDuckman]

@zhukovgreen Yes the manual reboot works.
Thanks

[zhukovgreen]

Thx @GeorgeDuckman We'll look into this and let you know

[zhukovgreen]

@fellipeh I see that this env={'LC_ALL': 'C'} in run_subprocess was introduced in #83. Any ideas around this issue?

[zhukovgreen]

Tracked under https://issues.redhat.com/browse/OAMG-4719

[zhukovgreen]

Hi @GeorgeDuckman, are you unable to extract the conversion logs from the unbootable machine (located at /var/log/convert2rhel/convert2rhel.log). Thank you

[GeorgeDuckman]

Hello @zhukovgreen, below the requested file, I had to delete some lines for professional reasons:

convert2rhel.log

Thank you.

[zhukovgreen]

@oamg/Convert2RHEL it seems @GeorgeDuckman hit a bug. From the log, I see that convert2rhel runs reboot command which doesn't exists on centos6.

[zhukovgreen]

Thanks, @GeorgeDuckman, we'll notify you of fixing this. However, it might unblock you, if you will try to run the convert2rhel utility without -r option:
-r, --restart Restart the system when it is successfully converted
to RHEL to boot the new RHEL kernel.
Please let me now if this manual reboot works for you

[zhukovgreen]

@GeorgeDuckman could you please check output on the centos6 host which is failing to convert:

echo $PATH
ls /sbin/reboot
ls /sbin/shutdown

We are trying to reproduce the error

[GeorgeDuckman]

@zhukovgreen I'll see to revert the host to a CentOS 6 server as soon as possible.

But if it helps, in my first post I said that the reboot command exists, I managed to use it from the command line. I then reproduced the code of the reboot function of convert2rhel and the reboot command is well executed by removing the "env={'LC_ALL': 'C'}" parameter part, which is surprising I admit.

[bocekm]

@GeorgeDuckman, when you try the conversion again, please try changing the line that runs the reboot (https://github.com/oamg/convert2rhel/blob/v0.20/convert2rhel/utils.py#L117) to run_subprocess("/sbin/reboot"), instead of removing the env.
I hope that can resolve the issue.

[bocekm]

I was able to reproduce the issue in the CentOS 6.10 vagrant box we use for testing and development with just using -r option. The above I recommended ("/sbin/reboot") fixed that.

[pirat89]

The explanation of the root cause is here: https://github.com/oamg/convert2rhel/pull/251/files#r616752752

[GeorgeDuckman]

Sorry for the wait.
Here is some more information :

[root@serv~]# whereis reboot
reboot: /sbin/reboot /usr/share/man/man2/reboot.2.gz /usr/share/man/man8/reboot.8.gz

[root@serv~]# whereis shutdown
shutdown: /sbin/shutdown /usr/share/man/man3p/shutdown.3p.gz /usr/share/man/man2/shutdown.2.gz /usr/share/man/man8/shutdown.8.gz

And thanks @bocekm and @pirat89 , I tried replacing "reboot" with "/sbin/reboot" and it works flawlessly now. Here's the tiny script I used to test it :

import subprocess
import sys
import shlex

def run_subprocess(cmd="", print_cmd=True, print_output=True):
    """Call the passed command and optionally log the called command (print_cmd=True) and its
    output (print_output=True). Switching off printing the command can be useful in case it contains
    a password in plain text.
    """
    if print_cmd:
        print("Calling command '%s'" % cmd)

    # Python 2.6 has a bug in shlex that interprets certain characters in a string as
    # a NULL character. This is a workaround that encodes the string to avoid the issue.
    if sys.version_info[0] == 2 and sys.version_info[1] == 6:
        cmd = cmd.encode("ascii")
    cmd = shlex.split(cmd, False)
    process = subprocess.Popen(cmd,
                               stdout=subprocess.PIPE,
                               stderr=subprocess.STDOUT,
                               bufsize=1,
                               env={'LC_ALL': 'C'})
    output = ''
    for line in iter(process.stdout.readline, b''):
        output += line.decode()
        if print_output:
            print(line.decode().rstrip('\n'))

    # Call communicate() to wait for the process to terminate so that we can get the return code by poll().
    # It's just for py2.6, py2.7+/3 doesn't need this.
    process.communicate()

    return_code = process.poll()
    return output, return_code

run_subprocess("/sbin/reboot")