Datapump files: how to analyse
This thread is about methods of analyzing the contents of Oracle data pump dump files.
In my company we have a support department routinely receiving export dump files from our customers that are imported locally in order to conduct reproduction of issues and providing support on configuration. These dump files can be anything like Oracle datapump dumps, old-style Oracle dump files or non-Oracle database dump files. An in-house tool conducts operations like importing, reloading and exporting. So far this tool cannot handle datapump dumps. I am currently defining specs for a new version of our tool in order to support data pump too.
This situation means that when a dumpfile X is offered for processing by our tooling the systems knows very little of the file. In fact not even if it is an Oracle dump file or not. In the era of Oracle only having traditional export available not that much of a problem: with the first few lines of a old-style Oracle dump being more or less readable text it was possible to conduct a quick test whether it is a Oracle dump or not. Largely regardless of the size of the dump file.
And now for Oracle data pump...
I am aware of the possibility of generating a sql file based on a datapump file (CONTENT=METADATA_ONLY option and pl/sql equivalent).
Multiple drawbacks:
#1: The result is a .sql file that needs to be interpreted too;
#2: The file must be copied to a database server; to a Oracle directory before conducting the test. For multiple GB files copying over the network takes time and creates network load.
#3: The database server doing the analysis must be of equal or higher version than the dump file. In case an Oracle 11 server is doing the analysis but it turns out to be an Oracle 10 dump it mus yest once more be copied over the network to the corresponding database server.
So what I would like to know is whether there is some kind of fixed header in a datapump export file that can be read and interpreted without the involvement of a database server. Even the possibility of getting basic property values would help. Like 'is it a datapump file or not' or 'what Oracle version did create the dump'.
I did search for this quite extensively as I would expect something like this to be available but without success so far.
Any ideas, anybody? If so please share.
Thanks,
Henk