«

One of my SPA-3000's seems to have developed a fault with its web
interface, and I'm wondering if anyone else has seen this. The fault
is intermittent and may be getting worse. It doesn't happen in the
basic user tabs, but does in the advanced admin tabs, and sometimes
in the basic admin tabs. I think it probably relates to the total
amount of data in all the tabs (they all seem to be passed to the
browser, not just the one currently visible, but the total size and
number of the tabs increases through to the advanced admin mode).

In IE, the fault just shows as "this page cannot be displayed".
With repeatedly hitting reload, after more than 10 times, it works.
In Mozilla, the page is displayed, but the tabs are non-functional
the tabbed screens are all displayed at once one after the other,
the "Submit all changes" form buttons are missing, and after a
further pause, Mozilla pops up a box "The document contained no data"
(which can happen when the browser asks for a page, and the server
closes the TCP connection without even returning an HTTP header,
although I haven't checked if that's the case here).

My guess is the SPA-3000 has started having difficulty returning
the larger web page responses, possibly due to some resource
problem. With my software engineer's hat on, I would suspect
it has leaked some memory and now doesn't have enough left to
work properly. It was working for about a day since it was last
switched on. A soft reset as done by submitting a configuration
change does not fix the problem. I am remote from the unit and
I can't currently get anyone to power cycle it, but that will be
the next thing to try. I can't easily tell at the moment if the
telephony and VoIP functions are still working since this issue
came to light. It has not logged anything to the debug syslog
server relating to this issue.

Anyway, just wondering is anyone has seen anything similar?

--
Andrew Gabriel

In article <42c72dda$0$38039$>,
(Andrew Gabriel) writes:

> I am remote from the unit and I can't currently get anyone
> to power cycle it, but that will be the next thing to try.

That didn't make any difference.

--
Andrew Gabriel

In uk.telecom.voip on 03 Jul 2005 00:14:18 GMT,
(Andrew Gabriel) wrote:

}Anyway, just wondering is anyone has seen anything similar?

Reduce the TCP window size to 16K

{R}

In article <>,
{R} <> writes:
> In uk.telecom.voip on 03 Jul 2005 00:14:18 GMT,
> (Andrew Gabriel) wrote:
>
> }Anyway, just wondering is anyone has seen anything similar?
>
> Reduce the TCP window size to 16K

Yes, I found it earlier today going through the packet traces.

It looks like a bug in the spa-3000's TCP stack, which goes into
a retransmit loop when the unacked data gets to somewhere around
38k (smells a bit like faulty signed/unsigned 16 bit arithmetic,
possibly tripping up at 32k). I knocked the window size down to
20k, and it then works. There's still a problem with the TCP
stack in the spa-3000 retransmitting when it shouldn't be, but at
least it no longer RST's the connection before it's got the whole
web page returned, even though it's sent about 4 times more data
than it should have.

BTW, upgrading to the current series 3 firmware makes no difference.
However, what I did find is that the syslog messages are broken in
the Series 3 firmware (PRI header is missing from the packet), so
I've gone back to Series 2 firmware, where the syslog packets are
correctly constructed.

--
Andrew Gabriel

In uk.telecom.voip on 04 Jul 2005 21:34:44 GMT,
(Andrew Gabriel) wrote:

}In article <>,
} {R} <> writes:
}>
}> Reduce the TCP window size to 16K
}
}Yes, I found it earlier today going through the packet traces.
}
}It looks like a bug in the spa-3000's TCP stack, which goes into
}a retransmit loop when the unacked data gets to somewhere around
}38k (smells a bit like faulty signed/unsigned 16 bit arithmetic,
}possibly tripping up at 32k). I knocked the window size down to
}20k, and it then works. There's still a problem with the TCP
}stack in the spa-3000 retransmitting when it shouldn't be, but at
}least it no longer RST's the connection before it's got the whole
}web page returned, even though it's sent about 4 times more data
}than it should have.
}
}BTW, upgrading to the current series 3 firmware makes no difference.
}However, what I did find is that the syslog messages are broken in
}the Series 3 firmware (PRI header is missing from the packet), so
}I've gone back to Series 2 firmware, where the syslog packets are
}correctly constructed.

Both bugs have been reported to Sipura.


{R}