I'm using Python/Pexpect to spawn an SSH session to multiple routers. The code will work for one router but then the output of session.before will get out of sync with some routers so that it will return the output from a previous sendline. This seems particularly the case when sending a blank line (sendline()). Anyone got any ideas? Any insight would be really appreciated.
Below is a sample of what I'm seeing:
ssh_session.sendline('sh version')
while (iresult==2):
iresult = ssh_session.expect(['>','#','--More--'],timeout=SESSION_TIMEOUT)
debug_print("execute_1 " + str(iresult))
debug_print("execute_bef " + ssh_session.before)
debug_print("execute_af " + ssh_session.after)
thisoutput = ssh_session.before
output += thisoutput
if(iresult==2):
debug_print("exec MORE")
ssh_session.send(" ")
else:
debug_print("exec: end loop")
for cmd in config_commands:
debug_print("------------------------------------------------\n")
debug_print ("running command " + cmd.strip() + "\n")
iresult=2
ssh_session.sendline(cmd.strip())
while (iresult==2):
iresult = ssh_session.expect([prompt+">",prompt+"#"," --More-- "],timeout=SESSION_TIMEOUT)
thisoutput = ssh_session.before
debug_print("execute_1 " + str(iresult))
debug_print("execute_af " + ssh_session.after)
debug_print("execute_bef " + thisoutput)
thisoutput = ssh_session.before
output += thisoutput
if(iresult==2):
debug_print("exec MORE")
ssh_session.send(" ")
else:
debug_print("exec: end loop")
I get this:
logged in
exec: sh version
execute_1 1
execute_bef
R9
execute_af #
exec: end loop
------------------------------------------------
running command config t
execute_1 1
execute_af #
execute_bef sh version
Cisco IOS Software, 1841 Software (C1841-IPBASEK9-M), Version 15.1(4)M4, RELEASE SOFTWARE (fc1)
Technical Support: http://www.cisco.com/techsupport...
I've run into this before with pexpect (and I'm trying to remember how I worked around it).
You can re-synchronize with the terminal session by sending a return and then expecting for the prompt in a loop. When the expect times out then you know that you are synchronized.
The root cause is probably that you are either:
Calling send without a match expect (because you don't care about the output)
Running a command that produces output but expecting for a pattern in the middle of that output and then not to next prompt that is at end of the output. One way to deal with this is to change your expect pattern to "(.+)PROMPT" - this will expect until the next prompt and capture all the output of the command sent (which you can parse in the next step).
I faced a similar problem. I tried waiting for the command to be printed on the screen and the sending enter.
I you want to execute say command 'cmd', then you do:
session.send(cmd)
index = session.expect([cmd, pexpect.TIMEOUT], 1)
session.send('\n')
index = session.expect([whatever you expect])
Worked for me.
I'm not sure this is the root of your problem, but it may be worth a try.
Something I've run into is that when you spawn a session that starts with or lands you in a shell, you have to deal with quirks of the TERM type (vt220, color-xterm, etc.). You will see characters used to move the cursor or change colors. The problem is almost guaranteed to show up with the prompt; the string you are looking for to identify the prompt appears twice because of how color changes are handled (the prompt is sent, then codes to backspace, change the color, then the prompt is sent again... but expect sees both instances of the prompt).
Here's something that handles this, guaranteed to be ugly, hacky, not very Pythonic, and functional:
import pexpect
# wait_for_prompt: handle terminal prompt craziness
# returns either the pexpect.before contents that occurred before the
# first sighting of the prompt, or returns False if we had a timeout
#
def wait_for_prompt(session, wait_for_this, wait_timeout=30):
status = session.expect([wait_for_this, pexpect.TIMEOUT, pexpect.EOF], timeout=wait_timeout)
if status != 0:
print 'ERROR : timeout waiting for "' + wait_for_this + '"'
return False
before = session.before # this is what we will want to return
# now look for and handle any additional sightings of the prompt
while True:
try:
session.expect(wait_for_this, timeout=0.1)
except:
# we expect a timeout here. All is normal. Move along, Citizen.
break # get out of the while loop
return before
s = pexpect.spawn('ssh me#myserver.local')
s.expect('password') # yes, we assume that the SSH key is already there
# and that we will successfully connect. I'm bad.
s.sendline('mypasswordisverysecure') # Also assuming the right password
prompt = 'me$'
wait_for_prompt(s, prompt)
s.sendline('df -h') # how full are my disks?
results = wait_for_prompt(s, prompt)
if results:
print results
sys.exit(0)
else:
print 'Misery. You lose.'
sys.exit(1)
I know this is an old thread, but I didn't find much about this online and I just got through making my own quick-and-dirty workaround for this. I'm also using pexpect to run through a list of network devices and record statistics and so forth, and my pexpect.spawn.before will also get out of sync sometimes. This happens very often on the faster, more modern devices for some reason.
My solution was to write an empty carriage return between each command, and check the len() of the .before variable. If it's too small, it means it only captured the prompt, which means it must be at least one command behind the actual ssh session. If that's the case, the program sends another empty line to move the actual data that I want into the .before variable:
def new_line(this, iteration):
if iteration > 4:
return data
else:
iteration+=1
this.expect(":")
this.sendline(" \r")
data = this.before
if len(data) < 50:
# The numer 50 was chosen because it should be longer than just the hostname and prompt of the device, but shorter than any actual output
data = new_line(this, iteration)
return data
def login(hostname):
this = pexpect.spawn("ssh %s" % hostname)
stop = this.expect([pexpect.TIMEOUT,pexpect.EOF,":"], timeout=20)
if stop == 2:
try:
this.sendline("\r")
this.expect(":")
this.sendline("show version\r")
version = new_line(this,0)
this.expect(":")
this.sendline("quit\r")
return version
except:
print 'failed to execute commands'
this.kill(0)
else:
print 'failed to login'
this.kill(0)
I accomplish this by a recursive command that will call itself until the .before variable finally captures the command's output, or until it calls itself 5 times, in which case it simply gives up.
Related
I'm writing a script (Python 2.7.10) to log in to networking devices and gather diagnostics information that the vendor can potentially ask for. It's fairly straightforward, but I've run into an interesting problem (at least for me).
I've exhausted my limited knowledge on this.
This is the piece of the code that calls the functions to run:
elif args.hostname and args.username and args.jtac and args.commands and args.shell:
print("RUN FOR SINGLE HOST w/ SHELL AND CLI COMMANDS")
open_ssh_session(args.hostname, args.username, password)
commands_and_iterations_cli(args.hostname, args.jtac, args.iterations, float(args.interval))
commands_and_iterations_shell(args.username, args.hostname, args.jtac, args.iterations, float(args.interval))
single_core_dump(args.hostname, args.username, password, args.jtac)
pull_files_from_juniper_device(args.hostname, args.username, password, args.jtac)
push_files_to_juniper_sftp(args.hostname, args.username, password, args.jtac)
So I have 2 functions:
def commands_and_iterations_cli(hostname, jtac, iterations, interval):
print("Enter each JUNOS CLI command on separate lines, press CTRL+D when finished.\n"
"NOTE: Valid CLI commands only, DO NOT input shell commands here!\n")
# Take user input, enter a command on each new line, send EOF to indicate that you're done.
cli_commands = sys.stdin.readlines()
# Instantiate user shell.
channel = client.invoke_shell()
# Take each line from given input, and iterate over hostname.
for line in cli_commands:
line = line.strip()
iter_counter = 0
print("Running {}, {} times, every {} seconds.".format(line, str(iterations), interval))
while iter_counter < iterations:
iter_counter = iter_counter + 1
channel.send(line +' | save "{}-{}-{}"\n'.format(hostname, jtac, line))
time.sleep(interval)
def commands_and_iterations_shell(username, hostname, jtac, iterations, interval):
print("Enter each shell command on separate lines, press CTRL+D when finished.\n"
"NOTE: Valid shell commands only, DO NOT input JUNOS CLI commands here!\n"
"** ENSURE COMMANDS ARE SAFE TO RUN IN PRODUCTION IF DOING SO! **\n")
# Take user input, enter a command on each new line, send EOF to indicate that you're done.
shell_commands = sys.stdin.readlines()
# Prompt for hostname's root password to enter root shell.
print("Enter root password for {}:\n".format(hostname))
rootpass = getpass.getpass()
# Instantiate user shell.
channel = client.invoke_shell()
# Start root shell.
channel.send("start shell user root\n")
# Let the prompt become available.
time.sleep(1)
# Send the password and let the root prompt return.
channel.send(rootpass+"\n")
time.sleep(1)
# Take each line from given input, and iterate over hostname.
for line in shell_commands:
line = line.strip()
print("Running {}, {} times, every {} seconds.".format(line, str(iterations), interval))
iter_counter = 0
while iter_counter != iterations:
channel.send(line +' >> "/var/home/{}/{}-{}-{}" \n'.format(username, hostname, jtac, line))
time.sleep(interval)
iter_counter = iter_counter + 1
print(iter_counter)
The code runs perfectly if I run commands_and_iterations_cli() or commands_and_iterations_shell() independently. However when I try both (the example given), the CLI function will run correctly, then when it comes time to run the shell function, the shell function will print the text in it, and prompt for the root password, and then immediately skip to the next function after it without prompting for commands. I've even tried giving it a 30 second sleep before the shell function runs, and the behavior just follows.
Thanks, all.
It doesn't work because you have to use CTRL+D to exit sys.stdin.readlines() which closes the stream and further calls to it do nothing (i.e. return empty list).
Essentially, i want what is in this thread: Output to console while preserving user input in ruby, but in Python. I have googled for quite a while, and found an ALMOST working solution, except that it blocked the main thread, as long as i wasn't typing anything in and pressing enter.
Some output of what i don't want to happen is here:
/raw:jtv!jtv#jtv.tmi.twitch.tv PRIVMSG #cobaltstreak :USERCOLOR ullr_son_of_sif #DAA520
Some example input of what i want is:
:jtv!jtv#jtv.tmi.twitch.tv PRIVMSG #cobaltstreak :USERCOLOR ullr_son_of_sif #DAA520
/raw
PRIV:jtv!jtv#jtv.tmi.twitch.tv PRIVMSG #cobaltstreak :SPECIALUSER nightbot subscriber
MSG #cobaltstreak :This shouldn't be here, but on the same line with /raw
This meaning, i want the bottom line of the console to preserve input, while outputting everything happening in the main thread without affecting input.
My current code is:
def console(q, m, lock):
while 1:
raw_input() # After pressing Enter you'll be in "input mode"
with lock:
i = raw_input('> ')
cmd = i.split(' ')[0]
msg = i.strip(cmd + ' ')
q.put(cmd)
m.put(msg)
if cmd == 'quit':
break
as well has:
cmd = cmd_queue.get()
msg = msg_queue.get()
action = cmd_actions.get(cmd)
if action is not None:
action(stdout_lock, msg)
Note the code above is the very first couple of lines in my while loop.
I am on Windows and using python 2.7.6
I have pexpect working, but I am having problems printing the output back from it. In my test script below, it creates the ssh connection, and then sends a sudo su -, then my password, and then sends a line that would require sudo access to do (I have also added p.interact() a few times to make sure it is at root). The problem I am having, is with returning the output of the commands I run. In the end I am wanting to run some top commands, and some du -h, and other(much more complex) space commands. But currently when it tries to print p.before, I get:
Traceback (most recent call last):
File "./ssh.py", line 37, in <module>
print p.before()
TypeError: 'str' object is not callable
Here is the script I am working from(edited to remove my pass and such)
#!/usr/bin/env python
import pexpect
import struct, fcntl, os, sys, signal
def sigwinch_passthrough (sig, data):
# Check for buggy platforms (see pexpect.setwinsize()).
if 'TIOCGWINSZ' in dir(termios):
TIOCGWINSZ = termios.TIOCGWINSZ
else:
TIOCGWINSZ = 1074295912 # assume
s = struct.pack ("HHHH", 0, 0, 0, 0)
a = struct.unpack ('HHHH', fcntl.ioctl(sys.stdout.fileno(), TIOCGWINSZ , s))
global global_pexpect_instance
global_pexpect_instance.setwinsize(a[0],a[1])
ssh_newkey = 'Are you sure you want to continue connecting'
p=pexpect.spawn('ssh user#localhost')
i=p.expect([ssh_newkey,'password:',pexpect.EOF,pexpect.TIMEOUT],1)
if i==0:
print "I say yes"
p.sendline('yes')
i=p.expect([ssh_newkey,'password:',pexpect.EOF])
if i==1:
print "I give password",
p.sendline("mypassword")
elif i==2:
print "I either got key or connection timeout"
pass
elif i==3: #timeout
pass
global global_pexpect_instance
global_pexpect_instance = p
p.sendline("sudo su -")
p.sendline("mypasswd")
p.sendline("mkdir /home/user/test")
print p.before
I am working off of this link: http://linux.byexamples.com/archives/346/python-how-to-access-ssh-with-pexpect/
Any help is much appreciated.
EDIT: As Armin Rigo pointed out below. I was calling to p.before as a function like p.before(). Stupid mistake on my part, as this explains why I was getting this error today, and not yesterday when I was trying this. After making that change to my script, and modifying the command being sent, print p.before, and no output is returned. Any other ways to return output from a sendline() command?
Use logfile, that logfile is store all output in terminal.use that example code:-
child = pexpect.spawn("ssh user#localhost")
child.logfile = open("/tmp/mylog", "w")
child.expect(".*assword:")
child.send("guest\r")
child.expect(".*\$ ")
child.sendline("python -V\r")
open the log file and see everything in terminals event
To fetch the complete output after sendline use child.read()
e.g.
cmd_resp = pexpect.spawnu(cmd) # for execution of the command
str_to_search = 'Please Enter The Password'
cmd_resp.sendline('yes') # for sending the input 'yes'
resp = cmd_resp.expect([str_to_search, 'password:', EOF], timeout=30) # fetch the output status
if resp == 1:
cmd_resp.sendline(password)
resp = cmd_resp.expect([str_to_search, 'outputString:', EOF], timeout=30)
print(cmd_resp.read()) # to fetch the complete output log
p.before is a string - not a function. To see the output you have to write
print p.before.
Hope this might help you
I was asked to simulate CLI with Python.
This is what I did
def somefunction(a,b):
//codes here
//consider some other functions too
print "--- StackOverFlow Shell ---"
while True:
user_input = raw_input("#> ")
splitit = user_input.split(" ")
if splitit[0] == "add":
firstNum = splitit[1]
sNum = splitit[2]
result = somefunction(firstNum, sNum)
print result
//consider some other elif blocks with "sub", "div", etc
else:
print "Invalid Command"
I do also check the length of the list, here "splitit" I will allow only 3 argumets, first will be the operation, and second and third are the arguments with which some functions are to be performed, in case the argument is more than 3, for that i do put a check.
Though Somehow I manage to make it work, but is there a better way to achieve the same?
Use python CMD Module:
Check few examples given on the below pages
http://docs.python.org/library/cmd.html # Support for line-oriented command interpreters
http://www.doughellmann.com/PyMOTW/cmd - # Create line-oriented command processors
prompt can be set to a string to be printed each time the user is asked for a new command.
intro is the “welcome” message printed at the start of the program.
eg:
import cmd
class HelloWorld(cmd.Cmd):
"""Simple command processor example."""
prompt = 'prompt: '
intro = "Simple command processor example."
You should check out the VTE lib:
http://earobinson.wordpress.com/2007/09/10/python-vteterminal-example/
It works really well and you can very easily customize its look. This is how easy it is:
# make terminal
terminal = vte.Terminal()
terminal.connect ("child-exited", lambda term: gtk.main_quit())
terminal.fork_command()
# put the terminal in a scrollable window
terminal_window = gtk.ScrolledWindow()
terminal_window.set_policy(gtk.POLICY_AUTOMATIC, gtk.POLICY_AUTOMATIC)
terminal_window.add(terminal)
Based on Chapter 12 of the OTP in Action book and Cesarini's book I wrote this Erlang code:
Erlang:
p(Param) ->
?DBG("Starting~n", []),
Cmd = "python test.py",
Port = open_port({spawn,Cmd}, [stream,{line, 1024}, exit_status]),
?DBG("Opened the port: ~w~n", [Port]),
Payload = term_to_binary(list_to_binary(integer_to_list(Param))),
erlang:port_command(Port, Payload),
?DBG("Sent command to port: ~w~n", [Payload]),
?DBG("Ready to receive results for command: ~w~n", [Payload]),
receive
{Port, {data, Data}} ->
?DBG("Received data: ~w~n", [Data]),
{result, Text} = binary_to_term(Data),
Blah = binary_to_list(Text),
io:format("~p~n", [Blah]);
Other ->
io:format("Unexpected data: ~p~n", [Other])
end.
Python:
import sys
def main():
while True:
line = sys.stdin.readline().strip()
if line == "stop-good":
return 0
elif line == "stop-bad":
return 1
sys.stdout.write("Python got ")
sys.stdout.write(line)
sys.stdout.write("\n")
sys.stdout.flush()
if __name__ == "__main__":
sys.exit(main())
The Erlang code suspends at the recieve clause - it never gets any message.
I have also checked Python from a regular Linux shell - it prints out every user input (1 - "Python got 1").
Where is the mistake here? Why doesn't my Erlang code get anything back?
There are two points:
make sure that Python does not buffer your output, try running python -u in open_port
using term_to_binary/1 and binary_to_term/1 won't work, since they assume that Python is able to encode/decode Erlang External Term Format, which does not seem to be the case. If you want to go this route, check out ErlPort
Does your Param contain the command limiter for Python? (in this case I assume newline, "\n"). Also, list_to_binary/1 and then a term_to_binary/1 feels kinda wrong. term_to_binary/1 directly (including the newline) should be sufficient.