Receiving cdmp_time stamp dump directories

[Former Member]

Hi,

We are having some issue with oracle logs/dumps files in our BW system. We have noticed that our BW system is creating cdmp* directory at below location.

/oracle/BWQ/saptrace/diag/rdbms/bwq/BWQ/trace

These cdmp* directories again contains other trace files which are not much useful.

These cdmp* directories are getting created of around 50+MB, which is making FS full.

We have found that these are created due to user processes, but not sure. And these can be purged with below command.

adrci> purge -age 4320 -type UTSCDMP   (delete dumps older than 3 days / 1440*3).

My requirement is how can we stop generating these dump directories, As this is happening only in our QA system, other systems are not affected with these dump files.

We have found below parameters which seems to be relevant but not getting proper details on it.

Kindly suggest if you have any thoughts on this.

FYI, SAP BW version is Nw 7.02 and Oracle version is 11g

Regards,

Kiraang.

[Former Member · Accepted solution]

Hi,

Finally I found the cause of the cdmp* generation directories. We were receiving some dumps related to index unusable state everyday.

So these SAP dumps are the root cause for generating cdmp* directories (dumps) at OS level.

Now there is a way I have deal this issue i.e. by following below procedure,

GO to SE37 and execute function module RSDU_TAB_UNIQUE_INDX_LOCAL, and provide table name. It will create index for that table.

We need to do it for all the tables and then there won't be any cdmp* dumps at OS level.

Regards,

Kiraang.

[ACE-SAP]

Hi

You should try a first to understand why these logs are generated.

Check if event 10046 has been activated (show parameter event)

Check in table DBA_ENABLED_TRACES if DBMS_MONITOR has been activated.

Brconnect should be able to cleanup traces (brconnect -u / -f cleanup) if you set parameter cleanup_ora_trace in BR*Tools configuration file (init<SID>.sap).

1551504 - Cleaning up Oracle audit files

1647272 - Minor functional enhancements in BR*Tools (4)

1. Deleting Oracle 11g trace files in other directories (implemented in BR*Tools 7.20 patch 20)

By default, the BRCONNECT function "cleanup" only deletes Oracle 11g trace files in the subdirectories of the following directory:

<diagnostic_dest>/diag/rdbms/<db_name>/<INST_NAME>

However, there are many other locations where Oracle 11g can write trace files. As they are specified dynamically, BRCONNECT does not take them into account by default.

To still enable you to clean up at selected additional locations, the following init<DBSID>.sap parameters were introduced.

cleanup_ora_dir = (<ora_trace_dir1>, <ora_trace_dir2>, ...)

cleanup_ora_sdir = (<ora_trace_sdir1>, <ora_trace_sdir2>, ...)

If the "cleanup_ora_dir" parameter is set, BRCONNECT deletes Oracle trace fields directly in the specified directories.

If the "cleanup_ora_sdir" parameter is set, oracle trace files are deleted in the subdirectories "alert", "cdump", "hm", "incident", "incpkg", "ir", "lck", "metadata", "sweep", "stage", and "trace" of the specified directories.

Regards

[Former Member]

Hi,

Thanks for the reply, this is for your information,

We need to know how can we stop generating this dump/log files as in other systems are getting this. And regarding clenaup we have found purge command (already mentioned before)

adrci> purge -age 4320 -type UTSCDMP

Kindly let us know if you can guide us on stopping this. We have check trace logs & everything is clean.

Thanks.

[ACE-SAP]

Ok, so no trace seems to be active... It's in fact quite strange.

With the 'bucket' word in the trace name we could think it might be related to CBO & histogram, but that's a pure guess.

Could you please post the content of one of these files, maybe what's inside could help to find what is creating it.

BTW it is not good to run an Oracle system without setting the required events.

Read note  1431798 - Oracle 11.2.0: Database Parameter Settings and set event parameter accordingly.

You could also find recommended values for event in the readme.html file of the applied bundle patch

Regards

[Former Member]

Hi,

Sorry for delay in my response, Here is the content of one of the bucket file.

Detail log file has been attached, Kindly check.

Thanks.

bucket trace file.txt.zip

[ACE-SAP]

Hi

My guess about bucket was stupid, the term is used in trace files and is not linked to histogram.

Check if sql trace is active, it is deprecated (show parameter sql_trace) and not traced in table DBA_ENABLED_TRACES

Have you checked alert logs for errors that could cause the dumps you get in cdump dir ?

Have you tried a basic  purge -age 4320 -type CDUMP

Regards

[Former Member]

Hi,

sql_trace parameter is also off

And regarding purge I already have the command which I am using (mentioned in my previous messages), I want to stop the cdmp* directories from getting produced.

I will now also raise this query to SAP, and will update if I will get the resolution.

Thanks.