oracledump
Oracle dump to SQL from an oradata folder for a CI/CD pipeline
Published: Friday, Feb 10, 2023 Last modified: Thursday, Nov 14, 2024
As a devops engineer, I work with CI/CD pipelines. Creating, maintaining, fixing and speeding them up.
When I heard of a job that takes 30 minutes, I had to investigate and I found:
- oradata.tar.gz ~ 1.2GB, extracted to 8GB
- an old oracle version (no other would work with this), was used to bootstrap with it to run tests
Can it bootstrap faster and save developers a chunk of time?
Oracle containers
A quick google reveals https://github.com/gvenzl/oci-oracle-xe#image-flavors with faststart for faster database startup times.
Great, but now I need to convert oradata to SQL for /container-entrypoint-initdb.d.
There is no Oracle dump
A rediculous amount of time was lost with trial and error on Oracle’s frankly broken CLI tools upon https://github.com/kaihendry/oracle-sql/blob/main/Makefile
- sqlplus - can’t export to sqlfile
- impdp - can’t export data to sqlfile
- sqlcl - can only go table by table
Unloading a table to Insert statements is hugely problematic because of foreign key constraints. There is a work around to disable the constraints table by table and then put them back once all the data is imported. For example:
ALTER TABLE sales MODIFY CONSTRAINT sales_promo_fk DISABLE;
ALTER TABLE sales MODIFY CONSTRAINT sales_promo_fk ENABLE;
Non-CLI GUI options
I found dbeaver and datagrip both broken with Oracle.
Your best bet is Oracle SQL Developer.
Despite Oracle’s SQL developer being slow, unstable and limited wrt how many rows it can export; it was the best export to SQL tool. To make it workable I had to ensure no table had more than a 1000 rows of test data.
The next issue is service name which is omitted in the export. Prepend:
ALTER SESSION SET CONTAINER=XEPDB1
Next issue; my users were not setup. Usually there is admin/user paradigm, so mine look something like:
CREATE USER eg_development_owner IDENTIFIED BY eg_development_owner QUOTA UNLIMITED ON USERS;
GRANT CONNECT, CREATE SYNONYM, RESOURCE TO eg_development_owner;
CREATE USER eg_development IDENTIFIED BY eg_development QUOTA UNLIMITED ON USERS;
GRANT CONNECT, RESOURCE TO eg_development;
GRANT CREATE SYNONYM to eg_development;
My biggest stumbling block was not being able to login with the ORACLE_PASSWORD. You need to set the role as SYSDBA on their client!
prime.sql
Now I have a ~1M bootstrap /container-entrypoint-initdb.d/prime.sql
down from
a ~8GB oradata folder. It takes about four minutes to bootstrap currently to a
DATABASE IS READY TO USE! state.
Do have a look at my Oracle CLI experiments and do please offer any suggestions.