Re: ORA-3135 Fixed by PC Restart

From: Tim Gorman <tim_at_evdbt.com>
Date: Thu, 03 Oct 2013 11:32:13 -0600
Message-ID: <524DAA1D.3010405_at_evdbt.com>

Charlotte,
Great troubleshooting! To summarize:

• no 10046 traces created for these failed connections from this PC, so no apparent cause on the server
• no problems from other clients at the same time, so again no apparent cause on the server

So, focusing on this lone PC and not other PCs or the database server, might it be possible to compare the contents of the SQLNET.ORA and the TNSNAMES.ORA from this PC with other PCs with similar 11.2.0.1 client installations which are not having the problems? Some old listings on the Oracle forums point toward parameters in the SQLNET.ORA file or some syntax in the TNSNAMES.ORA file as potential issues, but they are close to 10 yrs old.

The other more scientific method might be to enable client-side TNS tracing through the SQLNET.ORA file, using...

    TRACE_LEVEL_CLIENT = admin
    TRACE_DIRECTORY_CLIENT = C:\TEMP
    TRACE_FILE_CLIENT = SQLNET.TRC
    TRACE_UNIQUE_CLIENT = ON

    LOG_DIRECTORY_CLIENT = C:\TEMP
    LOG_FILE_CLIENT = SQLNET.LOG Of course, "c:\temp" is just a suggestion, as are all the parameter values shown. And the trace output is of course nasty cryptic, but it might just emit a clue. Otherwise, it is good material to submit in an SR to Oracle Support, too.

Another alternative is always just to nuke this particular PC and re-install? But it seems like you've been having some fun with this conundrum, and it would be really cool to see it through to resolution.

Please let me know what you think?

Thanks!

-Tim

On 10/3/2013 10:34 AM, Charlotte Hammond wrote:
> Hi Tim / List,
> Â
> Thanks for your hints - I've got my SQL*Plus environment sorted now and it's producing the following error when the connection becomes impossible: ORA-28547: connection to server failed, probable Oracle Net admin error.  This occurs at the same time as ORA-03135 from SQL*Developer.
> Â
> Unfortunately we're not licenced to run ASH on this database but I enabled 10046 tracing for the whole database (not a problem as it's lightly loaded) and I don't see any trace files for the failing connections which suggests that it's not even getting as far as the database.Â
> Â
> I'm still inclined to think there's a problem on the client side.  Today when we started getting the ORA-03135s (and ORA-28547s) we did not restart the PC as we usually do. After about 3 hours the problem went away and we could connect again, so whatever is breaking seems to be transient.  During these 3 hours we could continue to connect to the database from other PCs suggesting no issues with the network, listener etc.
> Â
> Any other thoughts?!
> Â
> Many thanks,
> Charlotte
> Â
> ----- Forwarded Message -----
> From: Tim Gorman <tim_at_xxxxxxxxx>
> To: oracle-l_at_xxxxxxxxxxxxx
> Date: Wed, 02 Oct 2013 15:32:51 -0600
> Charlotte,
> The ORA-1041 indicates that there is something wrong with your Oracle
> client-side installation, or there is something wrong with the
> environment, because that error means that SQL*Plus cannot find the
> language-specific file to translate the error number (i.e. "-3135") into
> the corresponding descriptive text (i.e. "connection lost contact" in
> English).
> Which likely means that SQL*Plus is indeed experiencing the ORA-3135 as
> well, but not finding the ".msg" file is an additional problem on top of
> that.
> In the past, I've used a clue contained within MOS Note #730066.1
> (entitled "/Diagnosis of ORA-3135/3136 Timeouts When The Fault Is In The
> Database/") and queried the V$ACTIVE_SESSION_HISTORY view for the 20
> seconds preceding the occurrance of the ORA-3135 error up to 5 seconds
> afterwards. The exact time of the ORA-3135 error is easy to obtain
> because a corresponding ORA-3136 error is logged to the database
> alert.log file. While querying V$ACTIVE_SESSION_HISTORY, I look for
> wait-events (i.e. column EVENT) in the class of "Concurrency" or
> "Cluster" or anything other than "ON CPU" or "User I/O". A good
> starting query might be something like...
> select event, count(*) cnt
> from gv$active_session_history
> group by event
> having count(*) > 10
> order by cnt desc;
> ãEUREUR
> This might begin to point you towards the root cause, if it resides in
> the database. Please feel free to share the results of this query with
> the list, if possible?
> Thanks!
> -Tim
> ãEUREUR
> ãEUREUR
> On 10/2/2013 3:12 PM, Charlotte Hammond wrote:

>> Hi Tim / List,
>>
>> I'm getting ORA-03135 from JDBC (SQL*Developer v.3.0).  Using SQL*Plus
>> instead returns (when it fails) ORA-01041: internal error hostdef extension
>> doesn't exist.   The errors occur at the same time on SQL*Developer and
>> SQL*Plus so these appear to be different manifestations of the same problem.
>>
>> I've discounted a load problem on the database as other PC clients are not
>> having this problem, the server is practically idle, and the session fails to
>> connect after just 20 seconds, not the CONNECT_TIMEOUT value I specified (of
>> several minutes).   This, along with the fact that a PC restart fixes it, is
>> why I think the problem is on the client side.  But equally the client can
>> still connect to all other databases fine while having a problem with just
>> this one.   Which leaves me confused...!
>>
>> Charlotte
>>
>>
>> ----- Forwarded Message -----
>>
>> From: Tim Gorman <tim_at_xxxxxxxxx>
>>>> To: oracle-l_at_xxxxxxxxxxxxx
>>>> Date: Wed, 02 Oct 2013 13:21:18 -0600
>>>>
>>>>
>>>> Charlotte,
>>>>
>>>> Is this ever happening when connecting using SQL*Plus from the PC in
>>>> question?  Or is it only from an application?  If so, what is the
>>>> application using for connectivity?  Is it ODBC, JDBC, or a straight-out
>>>> Oracle OCI/TNS connection?
>>>>
>>>> The reason that the ORA-3135 happens is because there are a few
>>>> preliminary SQL statements that
>>    execute before the database connection
>>>> is
>>    considered "established", and delays in processing those SQL
>>>> statements might overrun the configured timeout.  Some of the causes of
>>>> the ORA-3135 originate within the database, but there is no reason why
>>>> it couldn't also be the client-side PC that is not responding
>>>> appropriately and promptly.  Hence the questions about the nature of the
>>>> client program.
>>>>
>>>> Thanks!
>>>>
>>>>
>>>>
>>>> ________________________________
>>>> From: Charlotte Hammond <charlottejanehammond_at_xxxxxxxxx>
>>>> To: ORACLE-L <oracle-l_at_xxxxxxxxxxxxx>
>>>> Sent: Wednesday, October 2, 2013 8:07 PM
>>>> Subject: ORA-3135 Fixed by PC Restart
>>>>
>>>>
>>>>
>>>> Hi All,
>>>>
>>>> I've got a problem with a Windows 7 PC client (11.2.0.1) which after a
>>>> period of time (days/hours) starts throwing ORA-03135 (after a hang of
>>>> about 20 seconds) when I try to connect to a particular database
>>>> (11.2.0.3).   Restarting the PC always fixes the problem for a while but is
>>>> not a convenient workaround.
>>>>
>>>> It only happens on this one PC and against one target database.  (i.e.
>>>> other PC's can connect to the problem DB, and also this PC can connect
>>>> happily to other databases).  I've traced the client and listener but not
>>>> seen anything helpful yet- just a 20 second gap in times on the client side
>>>> before it fails, and a listener trace that stops dead on the disconnect.
>>>> I've tried changing the CONNECT_TIMEOUT to no avail and gone through the
>>>> Oracle MOS doc on ORA-03135 troubleshooting.
>>>>
>>>> Just wondering if this sounded familiar to anyone?   Just seems the fact
>>>> that rebooting the PC fixes things suggests the problem lies there.  Both
>>>> PC and Server are very, very lightly loaded. (A bit vague, I know, but just
>>>> looking for guidance on where to look:  the client or the server?  Could it
>>>> be a firewall issue? - there is a firewall in the path - but don't know why
>>>> it would be intermittently failing to connect and the network guys aren't
>>>> very interested so I'd need something more than a vague guess to get
>>>> traction from them.  It's quite difficult to diagnose/fix when the problem
>>>> is not readily reproducible).
>>>>
>>>> Many thanks,
>>>> Charlotte
>>>>
>>>>
>> --
>> http://www.freelists.org/webpage/oracle-l
>>
>>
>>

> --
> http://www.freelists.org/webpage/oracle-l

>
>
>
>

--
http://www.freelists.org/webpage/oracle-l

Received on Thu Oct 03 2013 - 19:32:13 CEST