Hundreds of devices are often managed using the Paramiko tool. The main is to fetch files from the top every day, as a backup.
Today we found that the program has been running for 10 hours before it is over, just go up and see what happened.
Check the logs and find out where the files are on one server. Manually ssh login up, executed a LS command to get stuck,
It turned out to be a problem with this server's hard drive. No wonder you can't get a file.
But think about it, the program should be in a period of time to read the data out of time to exit the AH, how can be stuck there. Find the part that executed the command.
Sin, sout, serr = Ssh.exec_command (' tar-zc0/data/important-file.txt ')
This statement is past, but behind
Sout.read (10240)
The statement stuck there all the time and stopped moving. It seems that sout is not set to timeout. Check the source code of Paramiko. Look, in the client.py.
def exec_command (self, Command, bufsize=-1)
Chan = Self._transport.open_session ()
Chan.exec_command (command)
stdin = Chan.makefile (' WB ', bufsize)
stdout = Chan.makefile (' RB ', bufsize)
stderr = Chan.makefile_stderr (' RB ', bufsize)
Return stdin, stdout, stderr
That stout is chan.makefile () out. Keep watching makefile.
def makefile (self, *params):
Return Channelfile (* ([Self] + list (params))
There is nothing to see here, focus on the Channelfile read method, the Read method is found. Channelfile inherited the Bufferedfile, as the document writes, this
Channelfile is a File-like object, so the _read method should be the actual read to call the method. Look at the code
def _read (self, size):
return Self.channel.recv (size)
Very simply, it is the Recv method of the channel (the channel is the first parameter of the Makefile, self, which is the Chan of Transport.open_session).
Or go back to the channel, recv method
def recv (self, nbytes):
Try
out = Self.in_buffer.read (nbytes, Self.timeout)
Except Pipetimeout, E:
Raise Socket.timeout ()
ACK = Self._check_add_window (len (out))
# No need to hold the channel lock when sending this
If ack > 0:
m = Message ()
M.add_byte (Chr (msg_channel_window_adjust))
M.add_int (Self.remote_chanid)
M.add_int (ACK)
Self.transport._send_user_message (M)
Return out
From the bold part can be seen, the original recv can be set to timeout, but Paramiko default is not set. OK, set the timeout to OK.
Modified the Paramiko code, run a bit, the program is not stuck in that place, in addition to the problem of the server, the rest are normally taken.
Modify the method as follows, modify the client.py Exec_command methods as follows
def exec_command (self, command, bufsize=-1, timeout = None):
Chan = Self._transport.open_session ()
If timeout is not None:
Chan.settimeout (Timeout)
Chan.exec_command (command)
stdin = Chan.makefile (' WB ', bufsize)
stdout = Chan.makefile (' RB ', bufsize)
stderr = Chan.makefile_stderr (' RB ', bufsize)
Return stdin, stdout, stderr
The blackbody portion is the added part. And in the place of the call, instead
Sin, sout, serr = Ssh.exec_command (' tar-zc0/data/important-file.txt ', timeout = 20.0)
Changes to the Python SSH tool Paramiko