Server stops accepting new connections

Openfire Openfire Support
1

We are having a couple of issues with the server, running on WinXP SP2.

After a period of time, it seems to stop accepting new connections from clients. And some clients that are connected seem to become “orphanned”, where the server still thinks they are connected, but when the client tries to reconnect they are told that their username is already in use.

All clients are using either Exodus (if on Windows) or Gaim (if on Linux).

The problem is solved by a quick stop and start of the server - then users can connect and reconnect for another couple of hours.

I’‘ve looked in the server logs but can’‘t find anything that looks significant (not that I am well equipped to tell). What’'s the best way to diagnose this issue?

thanks.

2

Poor form to reply to one’‘s own post, but here’'s some more info.

A successful connection to the server on our internal network seems to look like this:

Client:

<stream:stream to=“a190399” xmlns=“jabber:client” xmlns:stream=“http://etherx.jabber.org/streams” xml:lang=“en” version=“1.0” >

Server:

<?xml version=''1.0'' encoding=''utf-8''?>

Client:

alan

Server:

alan</ digest>

Client:

alanExodusremoved

Server:

Client:

Server:

****

The unsuccessful connections look like this (they don’'t get as far):

Client:

<stream:stream to=“a190399” xmlns=“jabber:client” xmlns:stream=“http://etherx.jabber.org/streams” xml:lang=“en” version=“1.0” >

Server:

<?xml version=''1.0'' encoding=''utf-8''?>

Client:

alan

And that’‘s as far as it goes. The client eventually times out waiting for the server response. And a restart of the server will fix the issue… for a while. Also, I’‘ve turned off the XP SP2 firewall in case that makes a difference, but it does not (I’'ve obviously already unblocked the appropriate ports).

Message was edited by: dubh (took out the password digest)

3

Thanks for the packet info. We’'re taking a look at this asap, and hopefully will have a fix in the next couple days. Will update with more information tomorrow.

4

thanks - don’'t hesitate to contact if you need more diagnostic stuff.

cheers

5

We’'re adding better error handling to the next patch release which will be coming out today or tomorrow. Hopefully, this will help us get a better understanding of the issue. Thanks for your patience

6

I run into the problem which has some resemblance with this one. System is JDK 1.4.2 and 1.5 on WinXP SP1. Client is custom app, Smack is used as client xmpp lib, but there are many iq extensions + geoloc (JEP 80) implementation. Messenger is running with HSQLDB.

What I noticed, is that offline messages are killing server. If client is not connected, and more then one message is sent, server stops accepting connections, and it looks that one user is disabled forever. Only help is to delete user and restart.

7

Hey Miroslav,

Does this problem happen every time you follow these steps? Before I try to reproduce this case I have a couple of questions.

What I noticed, is that offline messages are killing

server. If client is not connected, and more then

one message is sent, server stops accepting

connections,

Do you need to send the messages almost concurrently or it doesn’'t matter the delay between the messages?

and it looks that one user is disabled

forever.

Does the server refuse new connections to any user or only to the user that was offline while the messages were sent?

Only help is to delete user and restart.

Which user do you delete? I guess you are referring to the user that was offline and he/she cannot log in again.

Any info you can give me to reproduce the problem is greatly appreciated.

Thanks,

– Gato

8

Hi Gaston

I had this problem few days ago, so some details were not right. I tried tonight again to go through scenario and here are more details:

Sometimes I can do this even with one message. Period in between messages doesn’'t matter.

After this server is not responding. But sessions from previously connected clients are still on, even clients are gone. Server restart doesn’‘t help. And I was wrong about deleting users, that doesn’'t help either. Interesting thing clients can connect, I can even create new user, but it chokes on Authentication, actually it times out. I would look there first.

Message is still in HSQLDB. Here is line from DB where offline message is added:

INSERT INTO JIVEOFFLINE VALUES(3,2,’‘001096517891470’’,364,’‘Alert43.65414-79.567725020.02004-09-30T00:18:11Z127’’)

and it was never removed.

This is how new user is added, one which can’'t be authenticated:

INSERT INTO JIVEUSERID VALUES(’‘mbosnjakovic’’,1,0,8)

INSERT INTO JIVEUSER VALUES(8,’‘welcome’’,’’’’,1,’’ ‘’,0,’‘001097817810183’’,’‘001097817810183’’)

Maybe you can see something wrong here, I can’'t?

Log files are completely clean. And Offline Message options are set as:

  • Store the Message

  • Always Store

Let me know if you need more information. Unfortunately now I have one completely dead server, need to figure out hot to get it back

Miroslav

9

Here is update on previous situation. After few restarts, server stated responding. Log files are clean.

I’'m wondering about one thing from log file. There is Warning entry there saying that Temp file is deleted:

15:05:53.089 WARN!! [main]org.mortbay.http.HttpContext.getTempDirectory(HttpContext.java:1254)23> Delete existing temp dir C:\DOCUME1\Miroslav\LOCALS1\Temp\Jetty__9090__ for WebApplicationContext[/,Jive Messenger]

Is it possible that problem is here, since server think that sessions are still on it keeps some files locked, and they can’'t be deleted and all this happens because server thinks it has clients connected and authenticated even they are actually trying to do that …

And after while locks are released, files normally deleted, all happy…

Miroslav

10

Was there a fix / patch ever released? The current version for download is still 1.1.0 from 9/28/04.

I am currently testing out JIVE, and it works awesome… well… when it does work. I have the same issue where new connections can not be established.

I am usin PSI client (for SSL).

When a new connection tries ot connect, the get the warning message about the SSL Cert, which they continue. At that point, it continually tries to establish a connection, but no luck.

As the original poster mentioned, if I restart JIVE, everythign is fine.

This issue is preventing me from providing this to the users.

Overall this is one of the best jabber software I have seen, and once this bug is resolved, I’'ll be happy

Ton

11

Was there a fix / patch ever released? The current

version for download is still 1.1.0 from 9/28/04.

I haven’‘t heard/seen anything. But theoretically the new V2 should be out soon… so maybe that’‘ll fix it. I hope so, as it’‘s impacting on my users’’ willingness to use IM.

Agree about the general coolness of Jive. I had a lot of trouble trying to get other Jabber servers running, and Jive just worked.

12

We’‘ve reproduced this issue with the 1.1 release, but it doesn’'t seem to be a problem with the 2.0 codebase. So, a fix is coming very soon!

-Matt

13

it doesn’'t seem to be a problem with the 2.0

codebase. So, a fix is coming very soon!

Fantastic! Thanks Matt.

Alan