Forum OpenACS Q&A: FTP logon problems

Posted by Bob OConnor on

This problem is driving me nuts!

I regularly use FTP to update my OpenACS applications on Linux 6.2 and Linux 7.0 remote boxes. I use Windows as my main terminal on a small LAN connected to DSL. I use WS_FTP pro as the client FTP program under windows. Much of the time, the login is fast, other times it times out or minutes later with a few retries it works. The ping times to the server are low ~120ms and the server is serving webpages and doing telnet....

During the Longgggg wait for WS_Ftp to connect, I can pull up a telnet window and connect with a FreeBSD server in another city and type FTP and connect to the OpenACS servers without delay.

Further, I can use WS_FTP locally on the Win machine to connect to the remote FreeBSD server quickly without problems.

It is also true that telnet logon to the OpenACS servers is slower than usual but within about 15 seconds it connects... Once connected, the connection is good and I see my command line keystrokes immediately... I just Checked TOP:

2:40pm  up 98 days, ....  Load is 0.39, 0.44, 0.48
86 processes: 85 sleeping, 1 running, 0 zombie, 0 stopped
CPU states:  2.7% user,  2.5% system,  0.0% nice, 94.7% idle

Here is what WS-FTP shows in the logs...

connecting to 123.456.123.45:21
Connected to 123.456.123.45 port 21
[160] from port 1195
...[here is where there is a long wait
    and it may or may not connect]

This *seems to be* worse in the afternoons and evenings and not a problem in the early morning. The same delays occur with the MS-DOS window and FTP.

It almost seems that the Linux box doesn't like FTP packets from windows boxes and likes packets from other Unix boxes.... but then, I might be paranoid.

Any hints, suggestions, or explainations are appreciated....


Posted by Don Baccus on
I have similar weirdnesses with my 6.2 boxes, sometime ftp to them works fine, othertimes they stall.  inetd does appear to be trying to start ftp, very strange.  This is linux-to-linux so I think your windows paranoia is being triggered by a handful of coincidences.

I've never bothered tracking it down since the box I have hooked up to my DSL line, and which I normally use to drop stuff for folks to download, or let folks in to send me stuff, is an old Pentium 200 running RH 5.2 (the box which ran the first demo installation of the bboard module running under Postgres, as a matter of fact!).

Ftp to that box is always reliable so I've never bothered trying to figure out the oddity with the 6.2 boxes (I can always use scp to copy  files from one to the other).

Posted by Sam Snow on
Sound like a DNS problem to me. A long wait on log-in is a clasic sign...

See number 17 on for more detail (and how to fix it).

Posted by Don Baccus on
Well, I have /etc/hosts set up and telnet doesn't take a long time, so  #17 in the article doesn't help me ...

The IPs of the local machines I try to access from are also in /etc/hosts.

Posted by Jon Griffin on
Actually it is a bug in glibc.

Yet another reason to not run 6.2. (yeah, yeah, don't start the argument again.)

Posted by Pascal Scheffers on
Erhm, install a newer glibc? It's fully backward compatible. Or just upgrade to 7.2.