Prev: OS_CPU_WAIT_TIME
Next: what characterset to use?
From: Robert on 9 Aug 2007 19:07
Oracle servers on Windows servers crash after converting to OID from TNS name resolution.

Open a command prompt window (note: I have not seen this in the Windows version of SQL Plus or TOAD. You must use a DOS command prompt window). Make sure the windows workstation has SQLNET.ORA configuration that will use OID for TNS name resolution.

Open a SQLPlus session to any database.

Run a script that issues multiple "CONNECT uname/pwd(a)some-alias" commands.

I use this environment often. I am running Windows XP on a laptop with an Oracle 10g client. At least once a day I will get a long series of messages that look like this:

sgslufread: Hard error on read, OS error = 10054
sgslufread: Hard error on read, OS error = 10054
sgslufread: Hard error on read, OS error = 10054
sgslufread: Hard error on read, OS error = 10054
sgslufread: Hard error on read, OS error = 10054


This error condition cannot be cleared short of CTRL-C to kill the SQL-Plus process. Microsoft MSDN and Google searches lead me to believe this is a WinSock error. MSDN says of this:

10054 WSAECONNRESET Connection reset by peer. An existing connection was forcibly closed by the remote host. This normally results if the peer application on the remote host is suddenly stopped, the host is rebooted, or the remote host uses a hard close (see setsockopt (Windows Sockets) for more information on the SO_LINGER option on the remote socket.) This error may also result if a connection was broken due to keep-alive activity detecting a failure while one or more operations are in progress. Operations that were in progress fail with WSAENETRESET. Subsequent operations fail with WSAECONNRESET.

Some Observations:

The database "hang" only occurs when database links are used.

The database server only uses OID when invoking a database link.

If OID is completely removed from the SQLNET.ORA you cannot make the server hang.

If OID is present you can always make it hang by forcing a name resolution using OID. Within half an hour of running multiple queries through database links the windows server will hang.

The WinSock error will not get written to any Oracle log file (it's a WinSock error, not an Oracle error).

The WinSock error will not get written to any spool file (ditto).

The WinSock error causes the client session to hang.

Windows database servers can be made more resilient by granting them more resources and giving Oracle more control over resource management.
From: sybrandb on 10 Aug 2007 01:24
On Thu, 9 Aug 2007 23:07:21 GMT, "Robert" <b126775(a)yahoo.com> wrote:

>Oracle servers on Windows servers crash after converting to OID from TNS name resolution.

No one can comment on this one, because you didn't post your Oracle
version nor your Winblows version.
Why do you think you don't need to do this?

--
Sybrand Bakker
Senior Oracle DBA
From: Frank van Bortel on 10 Aug 2007 05:26
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Robert wrote:
> Oracle servers on Windows servers crash after converting to OID from TNS
> name resolution.
>
>
> Open a command prompt window (note: I have not seen this in the
> Windows version of SQL Plus or TOAD. You must use a DOS command
> prompt window). Make sure the windows workstation has SQLNET.ORA
> configuration that will use OID for TNS name resolution.
> Open a SQLPlus session to any database.
> Run a script that issues multiple "CONNECT uname/pwd(a)some-alias"
> commands.
> I use this environment often. I am running Windows XP on a laptop
> with an Oracle 10g client. At least once a day I will get a long
> series of messages that look like this:
>
> sgslufread: Hard error on read, OS error = 10054
> sgslufread: Hard error on read, OS error = 10054
> sgslufread: Hard error on read, OS error = 10054
> sgslufread: Hard error on read, OS error = 10054
> sgslufread: Hard error on read, OS error = 10054
>
> This error condition cannot be cleared short of CTRL-C to kill the
> SQL-Plus process. Microsoft MSDN and Google searches lead me to believe
> this is a WinSock error. MSDN says of this:
>
> 10054 WSAECONNRESET Connection reset by peer. An existing connection was
> forcibly closed by the remote host. This normally results if the peer
> application on the remote host is suddenly stopped, the host is
> rebooted, or the remote host uses a hard close (see setsockopt (Windows
> Sockets) for more information on the SO_LINGER option on the remote
> socket.) This error may also result if a connection was broken due to
> keep-alive activity detecting a failure while one or more operations are
> in progress. Operations that were in progress fail with WSAENETRESET.
> Subsequent operations fail with WSAECONNRESET.
>
>
> Some Observations:
>
> The database "hang" only occurs when database links are used.
>
> The database server only uses OID when invoking a database link.
>
> If OID is completely removed from the SQLNET.ORA you cannot make the
> server hang.
>
> If OID is present you can always make it hang by forcing a name
> resolution using OID. Within half an hour of running multiple queries
> through database links the windows server will hang.
>
> The WinSock error will not get written to any Oracle log file (it's a
> WinSock error, not an Oracle error).
>
> The WinSock error will not get written to any spool file (ditto).
>
> The WinSock error causes the client session to hang.
>
> Windows database servers can be made more resilient by granting them
> more resources and giving Oracle more control over resource management.
>
Search Metaclunk for "sgslufread" - three hits.

One doc suggests the version of the client may be the problem;
patching 9iR2 and 10GR2 clients to the latest version would help.
10GR1 clients do not have this problem.

Post your versions.
- --
Regards,
Frank van Bortel

Top-posting is one way to shut me up...
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (MingW32)

iD8DBQFGvC9BLw8L4IAs830RAgwzAJ9m3z6GQPUvxJRtJwdo8P7yzfLjkQCdGIN5
EJ2GXE1IUb4iTXLi1Y5cLUY=
=pGQZ
-----END PGP SIGNATURE-----
From: Robert on 15 Aug 2007 11:32
Thank you Frank, I read the metalink articles. We are still looking. My
analysis at this point suggests a configuration problem but we have not
found the cause yet. We are using a tracing tool which shows "the virtual
circuit was terminated due to a time out or other failure". We are unsing
Oracle 9.2.0.6 on a Win2k host.


"Frank van Bortel" <frank.van.bortel(a)gmail.com> wrote in message
news:f9hfrd$o67$1(a)news4.zwoll1.ov.home.nl...
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Robert wrote:
>> Oracle servers on Windows servers crash after converting to OID from TNS
>> name resolution.
>>
>>
>> Open a command prompt window (note: I have not seen this in the
>> Windows version of SQL Plus or TOAD. You must use a DOS command
>> prompt window). Make sure the windows workstation has SQLNET.ORA
>> configuration that will use OID for TNS name resolution.
>> Open a SQLPlus session to any database.
>> Run a script that issues multiple "CONNECT uname/pwd(a)some-alias"
>> commands.
>> I use this environment often. I am running Windows XP on a laptop
>> with an Oracle 10g client. At least once a day I will get a long
>> series of messages that look like this:
>>
>> sgslufread: Hard error on read, OS error = 10054
>> sgslufread: Hard error on read, OS error = 10054
>> sgslufread: Hard error on read, OS error = 10054
>> sgslufread: Hard error on read, OS error = 10054
>> sgslufread: Hard error on read, OS error = 10054
>>
>> This error condition cannot be cleared short of CTRL-C to kill the
>> SQL-Plus process. Microsoft MSDN and Google searches lead me to believe
>> this is a WinSock error. MSDN says of this:
>>
>> 10054 WSAECONNRESET Connection reset by peer. An existing connection was
>> forcibly closed by the remote host. This normally results if the peer
>> application on the remote host is suddenly stopped, the host is
>> rebooted, or the remote host uses a hard close (see setsockopt (Windows
>> Sockets) for more information on the SO_LINGER option on the remote
>> socket.) This error may also result if a connection was broken due to
>> keep-alive activity detecting a failure while one or more operations are
>> in progress. Operations that were in progress fail with WSAENETRESET.
>> Subsequent operations fail with WSAECONNRESET.
>>
>>
>> Some Observations:
>>
>> The database "hang" only occurs when database links are used.
>>
>> The database server only uses OID when invoking a database link.
>>
>> If OID is completely removed from the SQLNET.ORA you cannot make the
>> server hang.
>>
>> If OID is present you can always make it hang by forcing a name
>> resolution using OID. Within half an hour of running multiple queries
>> through database links the windows server will hang.
>>
>> The WinSock error will not get written to any Oracle log file (it's a
>> WinSock error, not an Oracle error).
>>
>> The WinSock error will not get written to any spool file (ditto).
>>
>> The WinSock error causes the client session to hang.
>>
>> Windows database servers can be made more resilient by granting them
>> more resources and giving Oracle more control over resource management.
>>
> Search Metaclunk for "sgslufread" - three hits.
>
> One doc suggests the version of the client may be the problem;
> patching 9iR2 and 10GR2 clients to the latest version would help.
> 10GR1 clients do not have this problem.
>
> Post your versions.
> - --
> Regards,
> Frank van Bortel
>
> Top-posting is one way to shut me up...
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.1 (MingW32)
>
> iD8DBQFGvC9BLw8L4IAs830RAgwzAJ9m3z6GQPUvxJRtJwdo8P7yzfLjkQCdGIN5
> EJ2GXE1IUb4iTXLi1Y5cLUY=
> =pGQZ
> -----END PGP SIGNATURE-----


From: Robert on 22 Aug 2007 13:52
Update: We traced and tested this issue on 9i and 10g. The problem occurs in a 9.2.0.6 Oracle database when using dblinks. Oracle's fix is to roll back to TNSNAMS or upgrade to 10G. Oracle is not going to address the problem due to resource limitations.
"Robert" <b126775(a)yahoo.com> wrote in message news:JMJ489.MJC(a)news.boeing.com...
Oracle servers on Windows servers crash after converting to OID from TNS name resolution.

Open a command prompt window (note: I have not seen this in the Windows version of SQL Plus or TOAD. You must use a DOS command prompt window). Make sure the windows workstation has SQLNET.ORA configuration that will use OID for TNS name resolution.

Open a SQLPlus session to any database.

Run a script that issues multiple "CONNECT uname/pwd(a)some-alias" commands.

I use this environment often. I am running Windows XP on a laptop with an Oracle 10g client. At least once a day I will get a long series of messages that look like this:

sgslufread: Hard error on read, OS error = 10054
sgslufread: Hard error on read, OS error = 10054
sgslufread: Hard error on read, OS error = 10054
sgslufread: Hard error on read, OS error = 10054
sgslufread: Hard error on read, OS error = 10054


This error condition cannot be cleared short of CTRL-C to kill the SQL-Plus process. Microsoft MSDN and Google searches lead me to believe this is a WinSock error. MSDN says of this:

10054 WSAECONNRESET Connection reset by peer. An existing connection was forcibly closed by the remote host. This normally results if the peer application on the remote host is suddenly stopped, the host is rebooted, or the remote host uses a hard close (see setsockopt (Windows Sockets) for more information on the SO_LINGER option on the remote socket.) This error may also result if a connection was broken due to keep-alive activity detecting a failure while one or more operations are in progress. Operations that were in progress fail with WSAENETRESET. Subsequent operations fail with WSAECONNRESET.

Some Observations:

The database "hang" only occurs when database links are used.

The database server only uses OID when invoking a database link.

If OID is completely removed from the SQLNET.ORA you cannot make the server hang.

If OID is present you can always make it hang by forcing a name resolution using OID. Within half an hour of running multiple queries through database links the windows server will hang.

The WinSock error will not get written to any Oracle log file (it's a WinSock error, not an Oracle error).

The WinSock error will not get written to any spool file (ditto).

The WinSock error causes the client session to hang.

Windows database servers can be made more resilient by granting them more resources and giving Oracle more control over resource management.
 | 
Pages: 1
Prev: OS_CPU_WAIT_TIME
Next: what characterset to use?