Paramiko hangs while executing a large wget command

Hi I am having problems executing a command that performs a wget of a 100mb file over a Ubuntu 10 server. Shorter commands work fine except for this. The below class contains how I use paramiko and my different tries of overcoming this problem (see the different run or exec methods). In the case of exec_cmd the execution hangs on this line:

        out = self.in_buffer.read(nbytes, self.timeout)

from the recv method of the channel.py module from paramiko.

The same wget command works perfectly in a shell using the normal ssh utility from Mac.

"""
Management of SSH connections
"""

import logging
import os
import paramiko
import socket
import time
import StringIO


class SSHClient():
    def __init__(self):
        self._ssh_client = paramiko.SSHClient()
        self._ssh_client.load_system_host_keys()
        self._ssh_client.set_missing_host_key_policy(paramiko.AutoAddPolicy())
        self.time_out = 300
        self.wait = 5

    def connect(self, hostname, user, pkey):
        retry = self.time_out
        self.hostname = hostname
        logging.info("connecting to:%s user:%s key:%s" % (hostname, user, pkey))
        while retry > 0:
            try:
                self._ssh_client.connect(hostname,
                                         username=user,
                                         key_filename=os.path.expanduser(pkey),
                                  开发者_如何学编程       timeout=self.time_out)
                return
            except socket.error, (value,message):
                if value == 61 or value == 111:
                    logging.warning('SSH Connection refused, will retry in 5 seconds')
                    time.sleep(self.wait)
                    retry -= self.wait
                else:
                    raise
            except paramiko.BadHostKeyException:
                logging.warning("%s has an entry in ~/.ssh/known_hosts and it doesn't match" % self.server.hostname)
                logging.warning('Edit that file to remove the entry and then try again')
                retry = 0
            except EOFError:
                logging.warning('Unexpected Error from SSH Connection, retry in 5 seconds')
                time.sleep(self.wait)
                retry -= self.wait
        logging.error('Could not establish SSH connection')

    def exists(self, path):
        status = self.run('[ -a %s ] || echo "FALSE"' % path)
        if status[1].startswith('FALSE'):
            return 0
        return 1

    def shell(self):
        """
        Start an interactive shell session on the remote host.
        """
        channel = self._ssh_client.invoke_shell()
        interactive_shell(channel)

    def run(self, command):
        """
        Execute a command on the remote host.  Return a tuple containing
        an integer status and a string containing all output from the command.
        """
        logging.info('running:%s on %s' % (command, self.hostname))
        log_fp = StringIO.StringIO()
        status = 0
        try:
            t = self._ssh_client.exec_command(command)
        except paramiko.SSHException:
            logging.error("Error executing command: " + command)
            status = 1
        log_fp.write(t[1].read())
        log_fp.write(t[2].read())
        t[0].close()
        t[1].close()
        t[2].close()
        logging.info('output: %s' % log_fp.getvalue())
        return (status, log_fp.getvalue())

    def run_pty(self, command):
        """
        Execute a command on the remote host with a pseudo-terminal.
        Returns a string containing the output of the command.
        """
        logging.info('running:%s on %s' % (command, self.hostname))
        channel = self._ssh_client.get_transport().open_session()
        channel.get_pty()
        status = 0
        try:
            channel.exec_command(command)
        except:
            logging.error("Error executing command: " + command)
            status = 1
        return status, channel.recv(1024)

    def close(self):
        transport = self._ssh_client.get_transport()
        transport.close()

    def run_remote(self, cmd, check_exit_status=True, verbose=True, use_sudo=False):
        logging.info('running:%s on %s' % (cmd, self.hostname))
        ssh = self._ssh_client
        chan = ssh.get_transport().open_session()
        stdin = chan.makefile('wb')
        stdout = chan.makefile('rb')
        stderr = chan.makefile_stderr('rb')
        processed_cmd = cmd
        if use_sudo:
            processed_cmd = 'sudo -S bash -c "%s"' % cmd.replace('"', '\\"')
        chan.exec_command(processed_cmd)
        result = {
            'stdout': [],
            'stderr': [],
        }
        exit_status = chan.recv_exit_status()
        result['exit_status'] = exit_status

        def print_output():
            for line in stdout:
                result['stdout'].append(line)
                logging.info(line)
            for line in stderr:
                result['stderr'].append(line)
                logging.info(line)
        if verbose:
            print processed_cmd
            print_output()
        return exit_status,result 

    def exec_cmd(self, cmd):
        import select
        ssh = self._ssh_client
        channel = ssh.get_transport().open_session()
        END = "CMD_EPILOGqwkjidksjk58754dskhjdksjKDSL"
        cmd += ";echo " + END
        logging.info('running:%s on %s' % (cmd, self.hostname))
        channel.exec_command(cmd)
        out = ""
        buf = ""
        while END not in buf:
          rl, wl, xl = select.select([channel],[],[],0.0)
          if len(rl) > 0:
              # Must be stdout
              buf = channel.recv(1024)
              logging.info(buf)
              out += buf
        return 0, out

I had the same problem, my python script hung when the shell script I was running on a remote ssh client, did a wget command on a 400Mb file.

I found that adding a time-out to the wget command fixed the problem. Originally I had:

wget http://blah:8888/file.zip

now with this:

wget -q -T90 http://blah:8888/file.zip

it works like a charm!

Hope it helps.

1. In this case, I would go with list appending and then concatenation. Why? Well, strings are immutable in Python. That means that every time you use += you are basically creating two new strings and reading a third. If you create a list and append it, on the other hand, you halve the number of strings created.
2. Do you really need to call select multiple times? My understanding is that you don't really care if the process is thread-blocking. Since select is more or less a wrapper around the C method of the same name:

   select() and pselect() allow a program to monitor multiple file descriptors, waiting until one or more of the file descriptors become "ready" for some class of I/O operation (e.g., input possible). A file descriptor is con‐ sidered ready if it is possible to perform the corresponding I/O operation (e.g., read(2)) without blocking.

3. You are not listening for a socket.timeout Exception in your code.
4. Writing to stdout/the file system can be expensive, yet you are logging every single line which is returned by recv. Can you move the log line?
5. Have you considered handling reading the channel manually? The only code you technically need is:

try:
    out = self.in_buffer.read(nbytes, self.timeout)
except PipeTimeout, e:
    # do something with error

It isn't guaranteed, but it will cut out extra processing.