Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

"SERVER/process_packet: broken port, server exiting" errors

223 views
Skip to first unread message

Steve Flanagan

unread,
Oct 14, 2002, 10:24:50 AM10/14/02
to

We are getting this error from .5 to 3 times daily and have to backup/restore the database. This started when we began using a 802.11b Cisco Aironet RF network with 11 access points and 4 PCs with RF cards in cranes. The server is IBOpen 6.0.2.0 and the client is using Delphi6 with BDEv5.11. All PCs are Win2000: one crane has no SP, 3 are SP3, and the server is SP2.
Can you answer the following:

1. How/when does the interbase engine decide to declare a connection broken? Is it solely based on the loss of an associated TCP socket connection, or does it also consider delay over an open socket?
2. How does the database server handle abnormal disconnects? Does it clean up open transactions?

Thank you for your help.
- Steve Flanagan

Bill Todd

unread,
Oct 14, 2002, 2:27:07 PM10/14/02
to

On 14 Oct 2002 07:24:50 -0700, "Steve Flanagan" <sfla...@uss.com>
wrote:

>
>We are getting this error from .5 to 3 times daily and have to backup/restore the database. This started when we began using a 802.11b Cisco Aironet RF network with 11 access points and 4 PCs with RF cards in cranes. The server is IBOpen 6.0.2.0 and the client is using Delphi6 with BDEv5.11. All PCs are Win2000: one crane has no SP, 3 are SP3, and the server is SP2.
>Can you answer the following:
>
>1. How/when does the interbase engine decide to declare a connection broken? Is it solely based on the loss of an associated TCP socket connection, or does it also consider delay over an open socket?

If the server does not here from the client for a specified amount of
time it sends a packet to the client. If it sends several packets and
gets no reply it assumes the client is dead. You can change the time
interval in the IB configuration file.

>2. How does the database server handle abnormal disconnects? Does it clean up open transactions?

All active transactions for that attachment are rolled back.

>
>Thank you for your help.
>- Steve Flanagan


--
Bill (TeamB)
(TeamB cannot respond to questions received via email)

Steve Flanagan

unread,
Oct 14, 2002, 3:41:29 PM10/14/02
to

The ibconfig file does not have CONNECTION_TIMEOUT or
DUMMY_PACKET_INTERVAL entries so I assume it is using
the defaults of 180 secs and 60 secs as described in the OPGUIDE. Is this correct?

When it happens we find many user connections that
don't exist, usually 14-15
but have seen 22 when there should be only 10.
Is this the 'phantom connection' problem ?

Once we see the broken port error the database starts growing
and we have to backup/restore it since its performance degrades.
It is only 14Mb but it can grow to 30Mb in two days with
queries taking twice as long. So I don't think transactions
are being rolled back.

Thanks for the help.
- Steve Flanagan

>If the server does not here from the client for a specified amount of
>time it sends a packet to the client. If it sends several packets and
>gets no reply it assumes the client is dead. You can change the time
>interval in the IB configuration file.
>
>>2. How does the database server handle abnormal disconnects? Does it clean up open transactions?
>
>All active transactions for that attachment are rolled back.
>
>>

>

Bill Todd

unread,
Oct 14, 2002, 5:21:34 PM10/14/02
to

On 14 Oct 2002 12:41:29 -0700, "Steve Flanagan" <sfla...@uss.com>
wrote:

>


>The ibconfig file does not have CONNECTION_TIMEOUT or
>DUMMY_PACKET_INTERVAL entries so I assume it is using
>the defaults of 180 secs and 60 secs as described in the OPGUIDE. Is this correct?

Yep.

>
>When it happens we find many user connections that
>don't exist, usually 14-15
>but have seen 22 when there should be only 10.
>Is this the 'phantom connection' problem ?

Sounds like it.

>
>Once we see the broken port error the database starts growing
>and we have to backup/restore it since its performance degrades.
>It is only 14Mb but it can grow to 30Mb in two days with
>queries taking twice as long. So I don't think transactions
>are being rolled back.

If the users are loging in as someone other than SYSDBA or the
database owner try shutting down the database and running a sweep. If
they are logging in as SYSDBA or the owner shut down the IB server and
restart it then run a sweep. This will not reduce the DB size but it
will garbage collect the old record versions so the space can be
reused and it should solve your performance problem.

Obviously the best solution would be to correct the cause of the lost
connections.

Steve Flanagan

unread,
Oct 15, 2002, 2:49:47 PM10/15/02
to

We are seeing outages of 1-5 secs on the network, not 180 sec.

We have found it faster to run a backup/restore script, stop
ibguardian, swap database and restorefile names, and restart
than to do stop/restart/sweeps.

Is there any way to do additional logging, or to log what
the BDE is doing ?

Thanks for all the help.

-Steve Flanagan

Bill Todd

unread,
Oct 15, 2002, 3:13:56 PM10/15/02
to

On 15 Oct 2002 11:49:47 -0700, "Steve Flanagan" <sfla...@uss.com>
wrote:

>


>We are seeing outages of 1-5 secs on the network, not 180 sec.
>
>We have found it faster to run a backup/restore script, stop
>ibguardian, swap database and restorefile names, and restart
>than to do stop/restart/sweeps.
>
>Is there any way to do additional logging, or to log what
>the BDE is doing ?

Not in current versions of InterBase. In a paper at the last BorCon
Charlie Caro said that IB 7 will have substantial new monitoring
features implemented though temporary tables that you can query to see
what is going on.

>
>Thanks for all the help.
>
>-Steve Flanagan

0 new messages