Web Agent - MEDIAPUT ORA-3113

Question

Hi,
I'm running an 8.1.6 db and using WebAgent on Redhat 6.1.
I have a running application to upload files into blob columns that usually works well except that from time to time the following error message appears:
"Oracle interMedia Web Agent
An error occurred processing your MEDIAGET or MEDIAPUT request
Error while trying to retrieve text for error ORA-03113"
I have no idea why I'm getting this error, if I refresh the page then I get the normal MEDIAPUT succeeded message and the document is loaded.
This is a little disconcerting for my end-users. Does anyone know what is causing this error or where to start looking?
Many Thanks.
Niall.
null

Answer 1

Hi Niall,
Unfortunately, the fact that you're not seeing any trace files or
anything in the alert log isn't giving us much to go on. I'll answer
your questions first, then list some ideas at the end.
The error message you are seeing, "MWM-00608: error executing a SQL
statement", indicates that the Web Agent really is executing a
statement - its pretty specific about what its doing when it has to
write an error message. For example, if the error were occuring when
it was trying to connect to the server or start a new database
session, then you'd see something like "MWM-00546: OCI error attaching
to database server using service name '%s'" or "MWM-00548: OCI error
beginning database session using service name '%s'".
Are there any delay / timeout parameters that should be examined or
changed ?The Web Agent doesn't use any timers. It simply issues calls to OCI and
waits for the response.
Does the Web Agnet try to keep it's connection alive from the previous
upload ?Yes it does. For database agents defined with a fixed user name and
password, the Web Agent keeps both the server connection and database
session active for use in subsequest requests. For database agents
that do not specify a password, the Web Agent keeps only the server
connection active between requests. It does this to avoid the overhead
of creating a new server connection/database session for every
request, something which would result in a noticeable in servicing
requests.
The fact that the MEDIAPUT request works immediately after the error
(upon page refresh) makes me wonder if the server process is not
coming up fast enough for the Web Agent? A page refresh should result in the browser resending the same request
to the web server. When the web agent gets the request, it will simply
create a new server connection/database session with which to execute
the necessary SQL. That is, assuming the request goes to the same
process. If it goes to a different process, then there may already be
an existing connection that can be used, or a new connection may be
created.
Here are some thoughts as to what might be causing the problem, plus
some suggestions where I can make any:
1. The database is being shutdown and restarted without restarting the
web agent and/or web server. This will cause the error you are
seeing, because the Web Agent doesn't know the database has gone
down and will try to use existing sessions when new requests
arrive.
If you're using Apache, then you'll need to restart Apache in order
to restart the Web Agent. If you're using iPlanet Web Server in
single-process mode, then you can use the Web Agent's on-line admin
interface to reload the configuration, which will cause the Web
Agent to shutdown any existing database sessions and server
connections, then re-read the configuration file. As new requests
are received, new connections will be established.
2. Individual database sessions are being killed by something and/or
someone. Although some documentation I read says that clients
should get an ORA-00028 error message, when I tried it, I got
ORA-03113. I didn't see anything in my alert log when I was trying
it, but there may be settings you can use to enable the logging of
such actions.
3. Individual TCP/IP network connections are being killed by something
and/or someone. I don't know how you're web server and database
server are configured, or what transport they're using, but this
sort of thing will result in the error you are seeing.
4. There's some sort of bug in the 8.1.6 server on Linux which results
in the server consuming virtual memory or some other resource to the
point that when the resource runs out, it doesn't have enough resource
to write a trace dump file.
If this is happening, then its going to be a case for the support
folks to look at. One possible way of proving this would be to
periodically restart the web server or reload the Web Agent
configuration, so re-initializing all the connections.
Some other questions that may or may not have a bearing on the issue.
- How often does this occur? Once a day, once a week, or multiple
times per day to different users?
- Which web server are you using?
- Does it always happen to the empentblobs.empent_doc procedure or
do other procedures experience problems?
- What does the empentblobs.empent_doc procedure do, and can it be
simplified in any way to narrow down the problem?
That's all I can think of for now. Please let us know if any of this
helps, or at least helps to point in the right directory.
Regards,
Simon
null