<body xmlns='http://jabber.org/protocol/httpbind' type='terminate' condition='item-not-found'/>
750 views
Skip to first unread message
khl
Jul 2, 2009, 2:22:59 AM7/2/09
to Strophe
Why do I receive the above stanza from the server? Once I receive
that, the next POST that Strophe makes to the server will result in
the response "invalid SID". And the Strophe connection becomes dead
at that point.
Prior to that, I sent a custom message to MUC. The response I
received is then passed to a handler, which does not do anything
strange.
that, the next POST that Strophe makes to the server will result in
the response "invalid SID". And the Strophe connection becomes dead
at that point.
Prior to that, I sent a custom message to MUC. The response I
received is then passed to a handler, which does not do anything
strange.
Jack Moffitt
Jul 2, 2009, 10:36:11 AM7/2/09
to str...@googlegroups.com
> Why do I receive the above stanza from the server? Once I receive
> that, the next POST that Strophe makes to the server will result in
> the response "invalid SID". And the Strophe connection becomes dead
> at that point.
> that, the next POST that Strophe makes to the server will result in
> the response "invalid SID". And the Strophe connection becomes dead
> at that point.
Are you load balancing at all? I think we were seeing this with
mod_http_bind in ejabberd in a load balancing set up that we had
incorrectly configured. Are you also using mod_http_bind?
jack.
Myers Carpenter
Jul 3, 2009, 2:32:45 PM7/3/09
to Strophe
xmlns='http://jabber.org/protocol/httpbind' type='terminate'
condition='item-not-found'/>" a 404 (which is what the old BOSH spec
said to do)".
Jack, How was your ejabberd server misconfigured? Did you not
replicate the http_bind mnesia tables to all the nodes?
myers
Jack Moffitt
Jul 3, 2009, 3:01:40 PM7/3/09
to str...@googlegroups.com
> Jack, How was your ejabberd server misconfigured? Did you not
> replicate the http_bind mnesia tables to all the nodes?
> replicate the http_bind mnesia tables to all the nodes?
I believe that was exactly what happened. Tofu may chime in with more
info if he remembers.
jack.
Christopher Zorn
Jul 6, 2009, 11:37:57 AM7/6/09
to str...@googlegroups.com
You can check if those are correct. Most likely they are. The problem we had was that there is a bug in http_bind that if the 'wait' is longer than 'inactivity' and there is a waiting request it will not wait for it but timeout.
After the timeout you will get the 404. You will need to check the logs, try replicating with a debug log level.
myers
Myers Carpenter
Jul 6, 2009, 1:56:38 PM7/6/09
to str...@googlegroups.com
On Mon, Jul 6, 2009 at 11:37 AM, Christopher
Zorn<christop...@gmail.com> wrote:
> You can check if those are correct. Most likely they are. The problem we had
> was that there is a bug in http_bind that if the 'wait' is longer than
> 'inactivity' and there is a waiting request it will not wait for it but
> timeout.
Zorn<christop...@gmail.com> wrote:
> You can check if those are correct. Most likely they are. The problem we had
> was that there is a bug in http_bind that if the 'wait' is longer than
> 'inactivity' and there is a waiting request it will not wait for it but
> timeout.
Wow, this could be a big break on the problem I'm currently working on.
Currently strophe.js on my site opens with this:
<body xmlns='http://jabber.org/protocol/httpbind'
sid='c111842b232e2130b0b0a4c69ab23a45c99ba86b'
wait='60' requests='2' inactivity='30' maxpause='120'
...
So if I understand what you are saying if I make wait="30" I will
avoid this bug?
A quick search for "inactivity" on the ejabberd ticket system didn't
turn up anything. Do you know if they have a ticket on this?
myers
Christopher Zorn
Jul 6, 2009, 4:37:41 PM7/6/09
to str...@googlegroups.com
On Mon, Jul 6, 2009 at 1:56 PM, Myers Carpenter <myers.c...@gmail.com> wrote:
Wow, this could be a big break on the problem I'm currently working on.
On Mon, Jul 6, 2009 at 11:37 AM, Christopher
Zorn<christop...@gmail.com> wrote:
> You can check if those are correct. Most likely they are. The problem we had
> was that there is a bug in http_bind that if the 'wait' is longer than
> 'inactivity' and there is a waiting request it will not wait for it but
> timeout.
Currently strophe.js on my site opens with this:
sid='c111842b232e2130b0b0a4c69ab23a45c99ba86b'
wait='60' requests='2' inactivity='30' maxpause='120'
...
So if I understand what you are saying if I make wait="30" I will
avoid this bug?
That could help. Bumping up inactivity will fix it. You can bump up the max inactivity value in the code too.
A quick search for "inactivity" on the ejabberd ticket system didn't
turn up anything. Do you know if they have a ticket on this?
They do not have a ticket but they are aware of it. We should create a ticket.
myers
Reply all
Reply to author
Forward
0 new messages