Does the morning report really need a login to your Oracle database?
The sales team needs yesterday’s order totals by 7 a.m. Somewhere along the way, that requirement became a service account, a firewall rule and a BI server that can open a session on the production Oracle database. The report runs for two minutes. The access path is there all day.
Oracle’s 17 November 2026 Critical Security Patch Update is a good reason to revisit that decision. Oracle released 655 new security patches across its product families. Three critical vulnerabilities stand out for anyone who has ever approved a “read-only” reporting account. CVE-2026-114819 and CVE-2026-114848 affect Oracle Text. CVE-2026-114822 affects the Oracle RDBMS. Oracle rates all three 9.9 on the CVSS 3.1 scale and lists CREATE SESSION as the required privilege on affected versions. A reporting account that logs in to run a SELECT already has that privilege. Oracle November 2026 advisory and risk matrices
The same advisory lists CVE-2026-114849 in both Oracle APEX and Oracle REST Data Services. Oracle rates it 9.8 and lists it as remotely exploitable without authentication. That should prompt a DBA to look at the other end of every approved connection, too. What software runs on the host that holds the reporting credentials, and who can patch it? Oracle November 2026 risk matrices
A BI gateway, an ETL server and a vendor-managed reporting application each bring their own drivers, runtimes and dependencies. Taken together, they may have more open CVEs than the Oracle installation they connect to. As a DBA, I can verify the patch level of my database. I may have no way to tell which driver a supplier bundled into its reporting product, whether a fix has shipped, or when the team operating it can deploy it. Even after a patch becomes available, my only option may be to open a ticket and wait. The reporting account and its firewall rule stay active in the meantime.
I can patch Oracle promptly and still be waiting for a supplier to fix its reporting tool. If someone compromises that tool in the meantime, they may be able to read production data using the reporting account’s existing permissions. That can happen even when Oracle is fully patched. If the report only needs yesterday’s totals, why give the tool a permanent route to production?
Let the database team publish the data
I’d start with that morning report. Which figures does Sales need? When does the business day close? What should the dashboard show if today’s delivery is late?
Suppose the answer is yesterday’s totals by store, ready by seven. The DBA schedules a job on a host the database team controls. The job runs the reclaimDB binary against an application-consistent, read-only snapshot. reclaimDB extracts the supported Oracle tables. The rest of the pipeline calculates the store totals, reconciles them against the daily control total and publishes the finished dataset. The raw extract stays inside the DBA-controlled environment. How reclaimDB extracts data from Oracle snapshots
That’s the shift from pull to push. The BI team can still explore the figures it receives. After both teams verify the delivery, the DBA can retire the reporting account and the network team can close the route from BI to production. If the BI server is compromised later, the delivered figures are still at risk. The standing connection back to Oracle is gone.
Give each team a format it already uses
For Oracle datafile extraction, reclaimDB can write Parquet, CSV, JSON and Avro, among other formats. For the daily sales figures, I’d start with Parquet. One agreed dataset can serve several familiar workflows. reclaimDB export formats
- Databricks can query the Parquet files directly in cloud storage. Snowflake can load the same files from a stage into a warehouse table with
COPY INTO. Databricks Parquet documentation · Snowflake Parquet loading - DuckDB lets an analyst run SQL directly against a delivered
.parquetfile. That’s enough to investigate why one store’s total dropped on Tuesday. DuckDB Parquet queries - Power BI can import Parquet through Power Query from Azure Blob Storage or Azure Data Lake Storage Gen2. The report can refresh from the published dataset using access scoped to that location. Microsoft Power Query Parquet connector
If Accounting already imports a small CSV every morning, I’d keep delivering CSV. An application that processes one record at a time might prefer newline-delimited JSON. The receiving team can choose the format without reopening the question of database access.
Put the next request in the backlog
Next month, Sales wants the figures broken down by channel. The BI owner puts that request in the backlog with an example of the expected result. The DBA updates the job that calls reclaimDB, adjusts the aggregation and adds a check that the channel figures still add up to each store’s daily total. The script or Ansible playbook lives in Git, where a colleague can review the change before the next scheduled run.
This is data engineering work, and I think it belongs within reach of every Oracle DBA. reclaimDB handles extraction. The DBA owns the pipeline. AI can help adapt a playbook, write a reconciliation check or make a failed delivery safe to retry. The DBA brings the judgment that matters most here, including which source is authoritative and when yesterday’s figures are final.
The export also has to fit the delivery window. In our published NFS test, reclaimDB exported eight intact tables without LOBs directly from Oracle datafiles to CSV. It finished about four seconds ahead of the Oracle Data Pump export on the same host and storage, with matching row counts for all eight tables. For the morning report, I’d measure the whole job through to the moment the validated figures are ready for Sales. reclaimDB and Oracle Data Pump export benchmark
Patch the affected systems. Then pull up the firewall rules for one critical Oracle database and pick a single reporting connection. If that team received exactly the data it needed, on time, would anyone miss the login?