nnimap woes again

nnimap woes again

[Doug Alcorn]

I've started running gnus in the same xemacs session.  Since then I've
been having aproblem with nnimap hanging.  When I start gnus, nnimap
works fine.  For a while it works fine too.  Then it hangs the next
time I try to get new messages.  Here's the messages from the "
*Message-Log*"

Opening nnimap server...
imap: Connecting to mail.lathi.net...
imap: Opening IMAP connection with `ssh -x -C mail.lathi.net /usr/sbin/imapd'...

-- after a long time, I interrupt it with C-g --

imap: Connecting to mail.lathi.net...failed

-- so i try to exit gnus --
Process not open for writing: #<process "imap" pid 4016 state:exit>

in the *nnimap* buffer I can see:
* PREAUTH [0.0.0.21] IMAP4rev1 v12.264 server ready

and that's it.  If I completely exit xemacs (and say yes to the prompt
about Active processes like so:

imap         run		  *nnimap* 		 /dev/pts/6	 /bin/bash -c ssh -x -C mail.lathi.net /usr/sbin/imapd

) I can restart gnus clean and it works.
-- 
 (__) Doug Alcorn (mailto:doug@lathi.net http://www.lathi.net)
 oo / PGP 02B3 1E26 BCF2 9AAF 93F1  61D7 450C B264 3E63 D543
 |_/  If you're a capitalist and you have the best goods and they're
      free, you don't have to proselytize, you just have to wait. 

Re: nnimap woes again

[Kai Großjohann]

When nnimap hangs, can you kill the according Gnus server buffer?
Name might be ` *nnimap*' or ` *server foo **nnimap*' or similar, with
a leading space.

A subsequent restart should work fine.

Or maybe you just do `^' to go to the server buffer and `C'lose and
`O'pen the server again.  Or maybe you need to `D'eny it first?  Not
sure.

kai
-- 
~/.signature: No such file or directory

Re: nnimap woes again

[Doug Alcorn]

Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann) writes:

> When nnimap hangs, can you kill the according Gnus server buffer?
> Name might be ` *nnimap*' or ` *server foo **nnimap*' or similar,
> with a leading space.

I can kill the buffer; but it doesn't do anything to alleviate the
problem.

> Or maybe you just do `^' to go to the server buffer and `C'lose and
> `O'pen the server again.  Or maybe you need to `D'eny it first?  Not
> sure.

I've tried this.  It just produces the same interminable hang.

-- 
 (__) Doug Alcorn (mailto:doug@lathi.net http://www.lathi.net)
 oo / PGP 02B3 1E26 BCF2 9AAF 93F1  61D7 450C B264 3E63 D543
 |_/  If you're a capitalist and you have the best goods and they're
      free, you don't have to proselytize, you just have to wait. 

Re: nnimap woes again

[Kai Großjohann]

On 23 Jul 2001, Doug Alcorn wrote:

> Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann) writes:
> 
>> When nnimap hangs, can you kill the according Gnus server buffer?
>> Name might be ` *nnimap*' or ` *server foo **nnimap*' or similar,
>> with a leading space.
> 
> I can kill the buffer; but it doesn't do anything to alleviate the
> problem.

Hm.  Is the corresponding process also killed?  (M-x list-processes RET)

kai
-- 
~/.signature: No such file or directory

Re: nnimap woes again

[Amos Gouaux]

>>>>> On Tue, 24 Jul 2001 00:18:08 +0200,
>>>>> Kai Großjohann <Kai.Grossjohann@CS.Uni-Dortmund.DE> (kg) writes:

kg> On 23 Jul 2001, Doug Alcorn wrote:
>> Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann) writes:
>> 
>>> When nnimap hangs, can you kill the according Gnus server buffer?
>>> Name might be ` *nnimap*' or ` *server foo **nnimap*' or similar,
>>> with a leading space.
>> 
>> I can kill the buffer; but it doesn't do anything to alleviate the
>> problem.

kg> Hm.  Is the corresponding process also killed?  (M-x list-processes RET)

Occasionally I've noticed that the ssl session (via stunnel) to our
IMAP server will hang.  I then have to `pkill openssl' and then do a
`g' in my *Group* buffer.  Usually this will cause me to login again
via ssl and things are peachy.

Just recently I noticed stunnel-3.15 was released, and from what I
read about it there was some mention of not handling timeouts well?
Was going to try that and see if it helped, that is until we can get
ourselves upgraded to Cyrus 2.X.

-- 
Amos

Re: nnimap woes again

[Doug Alcorn]

Amos Gouaux <amos+lists.ding@utdallas.edu> writes:

> Occasionally I've noticed that the ssl session (via stunnel) to our
> IMAP server will hang.  I then have to `pkill openssl' and then do a
> `g' in my *Group* buffer.  Usually this will cause me to login again
> via ssl and things are peachy.

I don't really know anything about stunnel.  I'm tunneling through ssh
using 'ssh -x -C mail.host /usr/sbin/imap'.  Would it be better/easier
if I used stunnel?  Using the ssh command, I can forego the login
issues with keys and a properly setup ssh-agent.
-- 
 (__) Doug Alcorn (mailto:doug@lathi.net http://www.lathi.net)
 oo / PGP 02B3 1E26 BCF2 9AAF 93F1  61D7 450C B264 3E63 D543
 |_/  If you're a capitalist and you have the best goods and they're
      free, you don't have to proselytize, you just have to wait. 

Re: nnimap woes again

[Amos Gouaux]

>>>>> On 24 Jul 2001 00:04:29 -0400,
>>>>> Doug Alcorn <doug@lathi.net> (da) writes:

da> I don't really know anything about stunnel.  I'm tunneling through ssh
da> using 'ssh -x -C mail.host /usr/sbin/imap'.  Would it be better/easier
da> if I used stunnel?  Using the ssh command, I can forego the login
da> issues with keys and a properly setup ssh-agent.

Probably not.  ssh-agent is pretty cool.

I'm just wondering if using a tunnel via ssh/ssl has its limits in
terms of reliable connectivity.  I'm wondering if STARTLS is the
only way to get a secure connection that is less likely to hang.
Dunno.

One thing we've noticed with OpenSSH 2.9X is that if a ssh session
dies due to a failed Internet connection, the process on the server
seems to hang around for almost indefinitely.  Maybe this is
related?  Dunno.

-- 
Amos

Re: nnimap woes again (more details)

[Doug Alcorn]

I got a suggestion for patching imap.el to put out some status
information (basically just reporting when it was looking for the
greeting).  So, I did the one line patch, byte-compiled and quit then
restarted gnus.  Right off the bat it hung.  I did a ps -ef to see the
if the process was there.
 
dalcorn   7055  7027  0 09:36 pts/6    00:00:00 ssh -x -C mail.lathi.net /usr/sbin/imapd
 
After waiting a while, I finally did a C-g since my xemacs session was
completely hung.  Immediately I did a ps -ef again to see the exact
same results as before.  IOW, the process hadn't died.  Then I did an
M-x list-processes and got this
 
imap         run		  *nnimap* 		 /dev/pts/6	 /bin/bash -c ssh -x -C mail.lathi.net /usr/sbin/imapd

The only buffer I had with the pattern "imap" in the name open was "
*nnimap*".  Here's it's contents.
 
* PREAUTH [0.0.0.21] IMAP4rev1 v12.264 server ready


Note that one of the interesting things is that I didn't restart
xemacs after I stopped gnus.  I just stopped gnus copied the elc file
and then restarted gnus.  Here's the sequence of messages:
 
Opening nnimap server...
imap: Connecting to mail.lathi.net...
imap: Opening IMAP connection with `ssh -x -C mail.lathi.net /usr/sbin/imapd'...
imap: Connecting to mail.lathi.net...failed

Also worth noting is that it never reported the message that gnus was
parsing the greeting.
-- 
 (__) Doug Alcorn (mailto:doug@lathi.net http://www.lathi.net)
 oo / PGP 02B3 1E26 BCF2 9AAF 93F1  61D7 450C B264 3E63 D543
 |_/  If you're a capitalist and you have the best goods and they're
      free, you don't have to proselytize, you just have to wait. 

Re: nnimap woes again

[Kai Großjohann]

On Mon, 23 Jul 2001, Amos Gouaux wrote:

> Occasionally I've noticed that the ssl session (via stunnel) to our
> IMAP server will hang.  I then have to `pkill openssl' and then do a
> `g' in my *Group* buffer.  Usually this will cause me to login again
> via ssl and things are peachy.

I've seen something similar with ssh: if the remote end drops the
connection, you don't see that until you type a command into your
local ssh client.  Only then does the local ssh client tell you that
the connection is down.

Gnus looks whether the connection is up before sending a command, but
once it has determined that the connection is up, it expects the
connection to be up and doesn't check again.

Maybe Gnus could be changed to see when it sent the last command.  If
that has been more than a configurable number of seconds ago, it sends
a no-op command to the remote end to see if the connection is still
up.  But nobody has implemented that, I'm afraid.

kai
-- 
~/.signature: No such file or directory

Re: nnimap woes again

[Christoph Rohland]

Hi Amos,

On Mon, 23 Jul 2001, Amos Gouaux wrote:
> Occasionally I've noticed that the ssl session (via stunnel) to our
> IMAP server will hang.  I then have to `pkill openssl' and then do a
> `g' in my *Group* buffer.  Usually this will cause me to login again
> via ssl and things are peachy.
> 
> Just recently I noticed stunnel-3.15 was released, and from what I
> read about it there was some mention of not handling timeouts well?
> Was going to try that and see if it helped, that is until we can get
> ourselves upgraded to Cyrus 2.X.

I also have problems running nnimap over openssl. But it is definitely
not timeout related. In my case most often copying between folders
hangs reproducable.

I switched of ssl because of that and it works fine without.

Greetings
		Christoph