Hi Helen,

Thanks for your comments.

The sites with this behaviour are running various OS's, including XP
Pro, XP Home and Win 98. All lanes are running Win 2k, except 1 running
XP - which is having regular problems. The test site which I am looking
to correct has the following: 7 x clients running 2k, 1 server running
Win XP Home SP1 + 2 other machines running 98. At most, this would
produce 15 connections to 3 databases on the server machine. All
machines are assigned static IPs when installed. This site does not
have an internet connection and so updates to the OS will be infrequent
to say the least. Standard windows name resolution is used (ie.
Servername:c:\databases\database.gdb to connect), and no machines are
set up in the hosts file.

In the case of the lockups, which is my main concern at the moment, it
seems to be directly related to the load on the machine, rather than a
firewall which cause it to fail all the time... I have introduced a few
logs and options to analyse the loss of events further, and will have
some results on that by Monday.

It is possible that the server does not have enough cache to handle the
requests (that is, I haven't looked into it), however the client will
still experience a lock up even though it is apparently not
communicating with the server. What seems to be the case is that
sometimes the load will simply slow down queries, which is acceptable
and expected, or sometimes some part of IBO (I assume) is hitting a
blocking call that is not serviced until the load is removed... I am
aware of Events having this behaviour, being in regular communication
with the server.

It is unlikely that there is a trojan running on the network.

Cheers.
Anthony

"Optimism is a cheerful frame of mind that enables a tea kettle to sing
though it's in hot water up to its nose."


-----Original Message-----
From: Helen Borrie [mailto:helebor@...]
Sent: Thursday, 2 September 2004 11:05 PM
To: isonaj77
Subject: Re: [IBO] IB_Events failing to fire


At 07:57 AM 2/09/2004 +0000, you wrote:

>--- In IBObjects@yahoogroups.com, Helen Borrie <helebor@t...> wrote:
>
> > >I believe a stray exception during the OnAlert event causes the
>IB_Event
> > >to stop firing. I have ensured there are no exceptions being
>raised
> > >when an Alert is triggered.
> >
> > Do you mean your code is deliberately swallowing exceptions?
>
>Yes, I have wrapped the EventAlert code in a single try..except and
>write the error to a log file in this case.
>
> > >Any ideas what may cause this strange behaviour?
> >
> > Installed WinXP SP2?
>
>This is something I will have to investigate.
>
> > Others (non IBO) are reporting event problems after installing
>SP2. Is
> > this in your equation?
> >
> > Is the client remote or local and, if local, how is it connecting?
>(protocol).
>
>Using a TCP/IP connection to a remote server as well as a TCP/IP
>connection to a local server. The events on the local server are
>working just fine.
>
> > If not this, what kinds of things have you done recently that might
>have
> > changed some conditions?
>
>The problem has increased over time, so the exact cause is difficult to

>identify. It does seem to be related to the number of clients connected

>(though our maximum is around 12 at this stage).
>
> > What database and version?
>
>Using Interbase 6.0.1.6
>
>
>There is a another problem we are experiencing that seems to be related

>to the problem above.
>
>First, a description of the system is needed:
>Many Pos clients are permanently connected to a local as well as a
>server database. The server database connection is not guaranteed to be

>reliable. Basic operations must be available at all times, whether the

>server is connected or not. Events are fired from both databases to
>initiate various functions. If there is a problem, the server is
>queried and if unavailable, disconnected. (eg. they power off the
>server) This is usually caught very quickly because the IB_Event blows

>up almost immediately. The second problem is this: If a hungry query is

>executed on the server, all clients lock up simultaneously, until that
>process completes. I cannot reproduce this problem except on a live
>site (of course :-( ) All inhouse tests show the clients to have a
>reduced throughput on server comms (as
>expected) but keep running. Server comms occur at known times and the
>lockup does not appear to be related to this, so I suspect the IB_Event

>to be the cause of this as well...
>
>Any hints or suggestions apreciated!

If both anomalies are coincidental with a Windows SP installation then
that's probably the source of the problems. Because the event "channel"

uses a random port pair for communication, a tightly locked firewall
(the
default bestowed with XP SP2) will lock off all of the unused ports.
Even
if XP SP2 isn't in the equation, check whether they installed or
reinstalled some form of firewall, e.g. Norton Internet Security, etc.

With Firebird, you can assign the event channel to a specific port but
with
IB 6 you'd be stumped unless you can have some unused ports open on the
server and one or two on each client. It's hypothetical, though, till
you
know their service pack history.

Is there any chance that there's a trojan running in the network?

Why is the server channel so unstable? Could it be that they are
running
DHCP and the server's IP address isn't declared in the Hosts files of
server and clients?

I assume you have enough server memory...? large enough temp sort space

configured? large enough sys memory cache on the server's disk?

Helen