Suddenly Contributor doesn´t work

[zeppomp3]

Planning 7.3
Contributor is installed in a server Windows 2003 SP1. This server is the job server, the web server.
there is only other server which is database server.

The problem is Contributor is working ok, but suddenly it hangs. And it´s neccesary restart the planning services. The issue ocurrs because there are many sockets in the contributor server in state CLOSE_WAIT produced by epPlanningService.exe. Where restart the service, this sockets Closed and all work fine.

the planningerrorlog.csv is "Unable to open connection using connection string of Provider=OraOLEDB.Oracle;Data Source=cpm;User ID=cpmuser;Password=*********; and timeout of 15 seconds.~~~~Error catastrófico"

the issue isn´t a Oracle Problem because the database server always works.
the services have a local administrator account.

Can I solve the problem?

Thanks.

[jluckhoff]

It think this could be a 2003 SP 1 thing:

Windows Server 2003 SP1 introduces a new security feature called Data Execution Prevention.  This prevents code from being compiled and stored in RAM on the server.  The Cognos Planning tool uses this technique in order to get the best possible performance when working with the planning model. 

This can cause unexpected Job Failures.  Windows kills the process completely so often there are no errors logged.

The setting is controlled in the windows system properties > advanced area.  Cognos planning requires that the option “Turn on DEP for essential windows programs and services only” is selected




Other Web Server Settings

1.1   MaxClientRequestBuffer
The MaxClientRequestBuffer limits the cumulative bytes size of the request line and header fields sent in a request to IIS.  The default value has been significantly reduced by Microsoft in latest versions of IIS and Windows.  This is to reduce the likelihood of an attack caused by requests containing large amounts of data in the request line or headers.

In the latest version of IIS (shipped with Win2003 and Win2K SP4) the default setting is just 16KB.  This can cause contributor web user requests to be rejected.  The default setting for IIS 4 was 2MB.
To increase or decrease the size of the buffer, simply navigate to the following registry key:

HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\w3svc\parameters

Then edit the MaxClientRequestBuffer key. If this key doesn't exist, add it, and set its data type to REG_DWORD. Then, in the DWORD Editor, select Decimal value, and then enter the number of bytes for the buffer.

1.2   AspMaxRequestEntityAllowed
This controls the amount of data you can upload to the server in an ASP request.  The default value for the AspMaxRequestEntityAllowed property is 204,800 bytes.  This may cause an error when saving or submitting data in contributor for larger applications.

To update this value you need to edit the MetaBase.  This should be done in conjunction with the customers local IT resource.


·   Stop the web service via Control Panel/Services.
·   Locate the file "MetaBase.xml which is by default in C:\Windows\System32\inetsrv.
·   Open it with Notepad.
·   Search for the property "AspMaxRequestEntityAllowed" and change its value (current recommendation is 1 GB 1073741824 bytes)
·   Start the web service.

1.3   AspBufferingLimit
This is the opposite of the AspMaxRequestEntityAllowed setting in that it controls the amount of data you are able to download from the server using ASP.  The default value for this in IIS 6.0 is 4MB and should be OK for most contributor models.  Again this parameter is set in the MetaBase and the suggested value if it is required to be increased is 8 MB (8388608 bytes)

Hope that works.

J

[zeppomp3]

Hello,

All these things that you say are ok.

[jluckhoff]

what version of oracle?

[zeppomp3]

the versions of oracle are the recommended versions by cognos Oracle Client 9.206 Oracle Server 9.204 and the patch.

[SomeClown]

Which release of 7.3?  You may want to consider upgrading to a more recent build if you are back several releases (Planning 8.1 is the same codebase as 7.3 so that is an option - it's an overlay and not uninstall/reinstall

Do you have a regularly scheduled reboot of the server?  This has been seen to be helpful at some implementations.

[jluckhoff]

Also check the state of the CEP COM+ components whenever the problem occurs. if the call time is really high and keeps going up, it would idicate a problem there. Probably best to reinstall the com objects at that point.

Also try to reconfigure your Job servers to use COM instead of HTTP to see if this clears it up - not a long term solution but at least it will give you some pointers.