Board index » delphi » events on a multihomed server - a bug or not a bug

events on a multihomed server - a bug or not a bug

// as posted also in borland.public.delphi.database.interbaseexpress
// but i feel like its more a general interbase issue in this case

Hi,

Discovered today : with IB6,  ibx4.6  ( ! jeff !) , server running NT4 ,
clients win98 , tcp/ip

Problem
- server goes down on a multihomed server (= 2 network cards) , when trying
to use ibevents on the client attached to the secondary network card.

Solution
had to make a clienside tcp ip route for all the secondary network attached
clients to the servers primary ip.

but IMHO , these problem is not tcp/ip related , but more IB server related.
thats also my question , is this is a bug or not a bug , thats the question.
and second, if this is not a bug, why would the ibserver crash because of an
unreachable client ?? and at last, could please somebody of the ibserver
development build in this feature of dealing with multiple server ip
addresses.

Situation

okay .. i have a server with 2 network cards.. so, ofcourse with 2 different
ip ranges.. lets say :
Network A : 10.0.0.x
Network B : 192.168.100.x

lets consider the server .. one of the 2 ip numbers assigned on the server,
is the primary ip.. lets say that network A is the primary network , so ..
doing a ping on the server itself to itselfs (ping ntserver) gives us (by
example) 10.0.0.2.

doing a 'ping ntserver' on network A client gives also 10.0.0.2 , but doing
a ping (resolve) on network B clients gives 'ping ntserver' 192.168.100.2
ofcourse because these clients are on the network B.

Working with an interbase database , something like ntserver:c:\database.gdb
.. is absolutly not a problem !! everything goes great as it should be ..

but... try using events in this situation.. Ofcourse the IB server needs to
remember all registerd 'events' clients.. so it probably remembers the ip
addresses where to send a event to .. but somehow the ibserver doesnt look
at the mulptiple ip addresses it has, and so, sends out 'events' /
'initalization stuff' to the clients on network B (192.168.100.x) but with a
reply address of its primary ip address (10.0.0.2) .. and not the ip address
that it should use for the client to be able to talk balk (192.168.100.2)..

thats why i had to make tcp/ip routes on all clients of the network B
(192.168.100.x)

for somebody who wants to know , i had to do :

Quote
>ROUTE ADD 10.0.0.0 MASK 255.255.255.0 192.168.100.2

okay.. maybe you say that this is a tcp/ip related issue... but it is still
strange that the server goes really down, isnit ? and second.. it just would
be a great future when ibserver is multihomed /multiple ip address aware,
isnt it ?

anybody

thanx

Lucas

 

Re:events on a multihomed server - a bug or not a bug


[please someone of teamb , please try this yourself !!!! ]
Hi,

Thanx for the reply.. but.. as told in the long story.. the crash of the IB
server only occurs when using IB events !!

so , ofcourse.. without the ip route and without the events , everything
works , users on side A can use interbase and users on side B can use the
same interbase server without any problems.

but but but.. when enabling ibevents in my program, without the route the
server crashes if i dont have setup a route , which shouldnt (in my opinion)
be necessary !! this is just because ibserver IS NOT multiple ip addresses
aware somehow..

please , please, someone.. test is on a machine

thanx lucas

Quote
"Jaimy Azle" <ja...@no-spam.plz> wrote in message news:3b40a14c_2@dnews...
> Hi,

> Unless you have install a firewall or setting any ip security, your
network
> B which on the eth1 (card #2) should be able to connect to net-1 gateway
> addr (10.0.0.1). as far as i know, ib just use the 1st address of your
> hostname to work. on a *n*x box, you can find isc_xxx.hostname file(s). i
> never find any problems using ib on the multiple ethernet card as i always
> use the 1st ip addr which is the hostname to work.

> > for somebody who wants to know , i had to do :
> > >ROUTE ADD 10.0.0.0 MASK 255.255.255.0 192.168.100.2

> you do not have to add this, unless you do want your networks to talk each
> others.

> HTH,

> Jaimy

Re:events on a multihomed server - a bug or not a bug


// as posted also in borland.public.delphi.database.interbaseexpress
// but i feel like its more a general interbase issue in this case

it IS a bug !!

see for yourself :

http://sourceforge.net/tracker/index.php?func=detail&aid=213460&group...
8&atid=109028

thanx to andres of finding this document !

everybody out there .. thats my bug !!!

please, please , could someone look at it.. if someone has questions about
this problem, you can contact for further details about it , cause i know
exactly when and why it occurs, but i just cant program that good to get
into the ibserver source...

Lucas Vos

Quote
"Lucas Vos" <v...@lucrasoft.nl> wrote in message news:3b3e70ea$1_1@dnews...
> // as posted also in borland.public.delphi.database.interbaseexpress
> // but i feel like its more a general interbase issue in this case

> Hi,

> Discovered today : with IB6,  ibx4.6  ( ! jeff !) , server running NT4 ,
> clients win98 , tcp/ip

> Problem
> - server goes down on a multihomed server (= 2 network cards) , when
trying
> to use ibevents on the client attached to the secondary network card.

> Solution
> had to make a clienside tcp ip route for all the secondary network
attached
> clients to the servers primary ip.

> but IMHO , these problem is not tcp/ip related , but more IB server
related.
> thats also my question , is this is a bug or not a bug , thats the
question.
> and second, if this is not a bug, why would the ibserver crash because of
an
> unreachable client ?? and at last, could please somebody of the ibserver
> development build in this feature of dealing with multiple server ip
> addresses.

> Situation

> okay .. i have a server with 2 network cards.. so, ofcourse with 2
different
> ip ranges.. lets say :
> Network A : 10.0.0.x
> Network B : 192.168.100.x

> lets consider the server .. one of the 2 ip numbers assigned on the
server,
> is the primary ip.. lets say that network A is the primary network , so ..
> doing a ping on the server itself to itselfs (ping ntserver) gives us (by
> example) 10.0.0.2.

> doing a 'ping ntserver' on network A client gives also 10.0.0.2 , but
doing
> a ping (resolve) on network B clients gives 'ping ntserver' 192.168.100.2
> ofcourse because these clients are on the network B.

> Working with an interbase database , something like

ntserver:c:\database.gdb

- Show quoted text -

Quote
> .. is absolutly not a problem !! everything goes great as it should be ..

> but... try using events in this situation.. Ofcourse the IB server needs
to
> remember all registerd 'events' clients.. so it probably remembers the ip
> addresses where to send a event to .. but somehow the ibserver doesnt look
> at the mulptiple ip addresses it has, and so, sends out 'events' /
> 'initalization stuff' to the clients on network B (192.168.100.x) but with
a
> reply address of its primary ip address (10.0.0.2) .. and not the ip
address
> that it should use for the client to be able to talk balk
(192.168.100.2)..

> thats why i had to make tcp/ip routes on all clients of the network B
> (192.168.100.x)

> for somebody who wants to know , i had to do :
> >ROUTE ADD 10.0.0.0 MASK 255.255.255.0 192.168.100.2

> okay.. maybe you say that this is a tcp/ip related issue... but it is
still
> strange that the server goes really down, isnit ? and second.. it just
would
> be a great future when ibserver is multihomed /multiple ip address aware,
> isnt it ?

> anybody

> thanx

> Lucas

Re:events on a multihomed server - a bug or not a bug


From SourceForge article
http://sourceforge.net/tracker/index.php?func=detail&aid=213460&group...
8&atid=109028

"Fix

By setting the internet IP first, then having the intranet ip set in the
advanced screen (under the internet IP), the problem is resolved (a reboot
will be required after changing the configuration). "

All you need to do is follow those instructions. If you follow those
instructions, and it DOESN'T fix it, I guess that's not the bug you have.

Quote
"Lucas Vos" <v...@lucrasoft.nl> wrote in message news:3b42597f_1@dnews...
> // as posted also in borland.public.delphi.database.interbaseexpress
> // but i feel like its more a general interbase issue in this case

> it IS a bug !!

> see for yourself :

http://sourceforge.net/tracker/index.php?func=detail&aid=213460&group...
Quote
> 8&atid=109028

> thanx to andres of finding this document !

> everybody out there .. thats my bug !!!

> please, please , could someone look at it.. if someone has questions about
> this problem, you can contact for further details about it , cause i know
> exactly when and why it occurs, but i just cant program that good to get
> into the ibserver source...

> Lucas Vos

> "Lucas Vos" <v...@lucrasoft.nl> wrote in message news:3b3e70ea$1_1@dnews...
> > // as posted also in borland.public.delphi.database.interbaseexpress
> > // but i feel like its more a general interbase issue in this case

> > Hi,

> > Discovered today : with IB6,  ibx4.6  ( ! jeff !) , server running NT4 ,
> > clients win98 , tcp/ip

> > Problem
> > - server goes down on a multihomed server (= 2 network cards) , when
> trying
> > to use ibevents on the client attached to the secondary network card.

> > Solution
> > had to make a clienside tcp ip route for all the secondary network
> attached
> > clients to the servers primary ip.

> > but IMHO , these problem is not tcp/ip related , but more IB server
> related.
> > thats also my question , is this is a bug or not a bug , thats the
> question.
> > and second, if this is not a bug, why would the ibserver crash because
of
> an
> > unreachable client ?? and at last, could please somebody of the ibserver
> > development build in this feature of dealing with multiple server ip
> > addresses.

> > Situation

> > okay .. i have a server with 2 network cards.. so, ofcourse with 2
> different
> > ip ranges.. lets say :
> > Network A : 10.0.0.x
> > Network B : 192.168.100.x

> > lets consider the server .. one of the 2 ip numbers assigned on the
> server,
> > is the primary ip.. lets say that network A is the primary network , so
..
> > doing a ping on the server itself to itselfs (ping ntserver) gives us
(by
> > example) 10.0.0.2.

> > doing a 'ping ntserver' on network A client gives also 10.0.0.2 , but
> doing
> > a ping (resolve) on network B clients gives 'ping ntserver'
192.168.100.2
> > ofcourse because these clients are on the network B.

> > Working with an interbase database , something like
> ntserver:c:\database.gdb
> > .. is absolutly not a problem !! everything goes great as it should be
..

> > but... try using events in this situation.. Ofcourse the IB server needs
> to
> > remember all registerd 'events' clients.. so it probably remembers the
ip
> > addresses where to send a event to .. but somehow the ibserver doesnt
look
> > at the mulptiple ip addresses it has, and so, sends out 'events' /
> > 'initalization stuff' to the clients on network B (192.168.100.x) but
with
> a
> > reply address of its primary ip address (10.0.0.2) .. and not the ip
> address
> > that it should use for the client to be able to talk balk
> (192.168.100.2)..

> > thats why i had to make tcp/ip routes on all clients of the network B
> > (192.168.100.x)

> > for somebody who wants to know , i had to do :
> > >ROUTE ADD 10.0.0.0 MASK 255.255.255.0 192.168.100.2

> > okay.. maybe you say that this is a tcp/ip related issue... but it is
> still
> > strange that the server goes really down, isnit ? and second.. it just
> would
> > be a great future when ibserver is multihomed /multiple ip address
aware,
> > isnt it ?

> > anybody

> > thanx

> > Lucas

Other Threads