SIGALRM problem

[Paul Rubin]

I'm trying to run a command on a remote host, something like:

result = os.popen('ssh otherhost mycommand').rea d()

It is possible that the other host is down, in which case the ssh
command hangs, so I want my script to time out if this happens:

def timeout(*x):
raise IOError, 'timeout'
signal(SIGALRM, timeout)
alarm(20)

result = os.popen('ssh otherhost mycommand').rea d()

I would expect the above to raise IOError if the ssh doesn't return
within 20 seconds, but it seems to hang when the other host is down,
until eventually the tcp connection times out after something like 5
minutes (which is much longer than I want to wait).

Anyone know if os.popen somehow pre-empts the alarm signal? If it
does, I'd consider that a bug.

Thanks
--Paul

[Mike Driscoll]

On Oct 13, 11:45*am, Paul Rubin <http://phr...@NOSPAM.i nvalidwrote:

I'm trying to run a command on a remote host, something like:

* *result = os.popen('ssh otherhost mycommand').rea d()

It is possible that the other host is down, in which case the ssh
command hangs, so I want my script to time out if this happens:

* *def timeout(*x):
* * * raise IOError, 'timeout'
* *signal(SIGALRM , timeout)
* *alarm(20)

* *result = os.popen('ssh otherhost mycommand').rea d()

I would expect the above to raise IOError if the ssh doesn't return
within 20 seconds, but it seems to hang when the other host is down,
until eventually the tcp connection times out after something like 5
minutes (which is much longer than I want to wait).

Anyone know if os.popen somehow pre-empts the alarm signal? *If it
does, I'd consider that a bug.

Thanks
--Paul

Using some Google-Fu, I found this post:

http://mail.python.org/pipermail/pyt...st/217530.html

It's old, but it looks ok...

Mike

[Paul Rubin]

Mike Driscoll <ky******@gmail .comwrites:

http://mail.python.org/pipermail/pyt...st/217530.html
It's old, but it looks ok...

Thanks. Maybe I did a dumb thing by overloading IOError and os.popen
is actually catching that exception for purposes of its own. I'll
try making a new exception.