ssh: don't use a separate logger

The logger setup was confusing when using --log-file-level=DEBUG,
showing this debug-level output inside default --log-cli-level=INFO output,
but without the debug context that only went into the logfile, and without
any hint of what kind of output it was.

This makes it use the same logger as all other debug output, preventing it
from appearing in the wrong place, and adds a marker to hint it is output
data - at the same time fixing a potential formatting issue, where
arbitrary ssh output lines were interpreted as *format strings*.

Signed-off-by: Yann Dirson <yann.dirson@vates.tech>

lib/commands.py

@@ -59,12 +59,6 @@ def _ellide_log_lines(log):
         reduced_message.append("(...)")
     return "\n{}".format("\n".join(reduced_message))
 
-OUPUT_LOGGER = logging.getLogger('output')
-OUPUT_LOGGER.propagate = False
-OUTPUT_HANDLER = logging.StreamHandler()
-OUPUT_LOGGER.addHandler(OUTPUT_HANDLER)
-OUTPUT_HANDLER.setFormatter(logging.Formatter('%(message)s'))
-
 def _ssh(hostname_or_ip, cmd, check, simple_output, suppress_fingerprint_warnings,
          background, target_os, decode, options):
     opts = list(options)
@@ -111,7 +105,7 @@ def _ssh(hostname_or_ip, cmd, check, simple_output, suppress_fingerprint_warning
     for line in iter(process.stdout.readline, b''):
         readable_line = line.decode(errors='replace').strip()
         stdout.append(line)
-        OUPUT_LOGGER.debug(readable_line)
+        logging.debug("> %s", readable_line)
     _, stderr = process.communicate()
     res = subprocess.CompletedProcess(ssh_cmd, process.returncode, b''.join(stdout), stderr)