outbound ssh stops working after some time/amount of data transferred

Hi experts,

I'm puzzled: I'm generating some files on my Solaris box and push them to another machine via ssh:

tar cvf - ${RESULT_FILE} \

| ssh ${TARGET_HOST}"cd ${TARGET_DIR}; tar xvf -"

The ssh is done via ssh-key, managed by keychain-2.6.2. Every three days this stops working and the ssh command gets stuck. This also happens when I ssh to another machine where I have to type the password.

Sometimes I happen to be logged in when this occurs, and then, within an hour, my X-session (KDE) freezes up.

A maybe related symptom is that on outbound rsyncs, via ssh

rsync -av -essh my_local_dir remote_host:/my_remote_dir

the data flow gets stuck as well.

I then log in from another machine, kill my session, come back, log in as root and try to check the log files of the services, etc., but they don't record the problem.

In general, I have a hard time finding out what's happening. Where to look?

Somehow it's stupid rebooting the machine every three days... It's so non-UNIX.

Cheers, Peter.

[1208 byte] By [niessepea] at [2007-11-26 12:40:45]
# 1

Can you describe the symptoms of "ssh stops working"?

Is the remote side still pingable?

What does 'ssh -v <host>' show?

Can you 'snoop' the network from either the client or server side to see data flow?

Can you log in to the target host another way (local console, telnet, etc. )while the problem is occuring?

--

Darren

Darren_Dunhama at 2007-7-7 16:12:45 > top of Java-index,General,Network Configurations...
# 2

> Can you describe the symptoms of "ssh stops

> working"?

Hi Darren,

thanks to the gallant knight who comes to the rescue!

I might add another thing: I'm authenticating my self through LDAP and my home directory is automounted from a (Linux) server.

>

> Is the remote side still pingable?

I guess, though I never tried this.

Some more symptoms:

-- I can still read and write to my home directory

-- When logging in as root, I can do ssh -l myname remotehost

-- I can still log onto the text console and ssh into the box.

-- However, starting an X-session will not work. The "Starting the K-Desktop" white screen with the solaris logo comes on, but KDE doesn't start up.

-- Maybe it's LDAP related. The pam.conf is such that root cannot start an X session, and su - doesn't accept the root password.

-- restarting the automount, sshd, ldap_cachemanager doesn't cure it. Maybe I do it in the wrong order?)

> What does 'ssh -v <host>' show?

the output of ssh -vvv can be seen at

http://forum.java.sun.com/thread.jspa?threadID=5099152&messageID=9339074#93 3907

> Can you 'snoop' the network from either the client or

> server side to see data flow?

Like with tcpdump on Linux?

> Can you log in to the target host another way (local

> console, telnet, etc. )while the problem is

> occuring?

Yes, that's usually no problem. When outbound ssh stops working, I sometimes log in from outside and suck the data out of the box instead of blowing it out.

I will be travelling for a while and not be able to access the machine for a while.

>

> --

> Darren

Once again, thanks for your time and effort.

Cheers, Peter.

niessepea at 2007-7-7 16:12:45 > top of Java-index,General,Network Configurations...
# 3

Nothing pops out at me immediately.

I would agree that it could be an authentication issue with LDAP. Trussing the remote (server) sshd during the hang might also be informative because you could see what system call it was hanging on. (I'll bet it's some sort of account lookup or similar).

--

Darren

Darren_Dunhama at 2007-7-7 16:12:45 > top of Java-index,General,Network Configurations...