The restart fixed it. The problem is still there.
An incident that closes without a reconstructed causal chain is not an isolated episode. How to read AWR, ASH and change logs after the fact before the second failure arrives.
Recent
Technical Leadership & Projects
AI shows the symptom. The board asks for the cause.
Datadog and Dynatrace show symptoms, not causes. How to explain to the board why modern observability isn't enough without a skilled reader.
Technical Leadership & Projects
Technical Leadership & Projects
More resources, same root cause
Why upgrading CPU, RAM or IOPS rarely solves a mission-critical database slowdown — and what to check before signing the purchase order.
MySQL
128 MB on a 32 GB machine: InnoDB buffer pool sizing in production
A MySQL 8.0 e-commerce with 32 GB RAM running on 128 MB buffer pool. How we diagnosed the issue, fixed it, and cut query latency from 45ms to 3ms.
Technical Leadership & Projects
The crazy doctor: a Commodore 64, a locked door, and the craft of understanding
A personal story about curiosity, a locked door, and a mad doctor program that changed everything. How a C64 in 1983 shaped thirty years of database work.
Oracle
The Saturday night nobody wants: migrating 12 TB from Oracle 12.2 to 21c in a four-hour window
Moving 12 TB from Oracle 12.2 to 21c with transportable tablespaces and RMAN incremental backups, keeping downtime under four hours. Real numbers and commands.
MySQL
The 8:40 AM phone call: recovering a Galera Cluster when nodes fall one after another
A Galera 3-node cluster loses quorum during a live incident. Diagnosis, emergency bootstrap, and recovery procedure guided over the phone — with the runbook.
Data Warehouse
Data Warehouse
ETL Oracle: from 4 hours to 25 minutes with staging tables, MERGE and parallel DML
A real Oracle 19c Data Warehouse case: how row-by-row inserts, missing indexes, and frequent COMMITs were killing a nightly ETL batch window.
Data Warehouse
The lunch break that delayed go-live: Data Governance in the DWH
A lunch conversation about what a technically ready Data Warehouse still lacks: data ownership, quality framework, GDPR, TDE on Oracle 19c.
Data Warehouse
Three data marts, three truths about sales: the bus matrix as common ground
Kimball bus matrix to align isolated data marts: conformed dimensions, business processes and comparable sales. Real insurance group case.
Data Warehouse
DWH Partitioning: When 3 Years of Data Weigh Too Much
Range partitioning on an 800-million-row fact table: from 12-minute quarterly queries to 40 seconds. Monthly implementation, exchange and local indexes.
Data Warehouse
Ragged hierarchies: when the client has no parent and the group has no grandparent
Ragged hierarchy in the data warehouse: balancing unbalanced trees with the self-parenting technique. Correct drill-down on customers and groups.
Data Warehouse
SCD Type 2: the history the business didn't know it needed
SCD Type 2 in the data warehouse: historicising dimensions with surrogate keys and validity dates. Real case: a customer dimension that evolves over time.
Data Warehouse
Wrong grain: when the fact table can't answer the right questions
Data warehouse: fact table grain determines which questions you can answer. Common mistakes in granularity and their impact on the dimensional model.
Technical Leadership & Projects
Technical Leadership & Projects
AI shows the symptom. The board asks for the cause.
Datadog and Dynatrace show symptoms, not causes. How to explain to the board why modern observability isn't enough without a skilled reader.
Technical Leadership & Projects
The restart fixed it. The problem is still there.
An incident that closes without a reconstructed causal chain is not an isolated episode. How to read AWR, ASH and change logs after the fact before the second failure arrives.
Technical Leadership & Projects
More resources, same root cause
Why upgrading CPU, RAM or IOPS rarely solves a mission-critical database slowdown — and what to check before signing the purchase order.
Technical Leadership & Projects
The crazy doctor: a Commodore 64, a locked door, and the craft of understanding
A personal story about curiosity, a locked door, and a mad doctor program that changed everything. How a C64 in 1983 shaped thirty years of database work.
Technical Leadership & Projects
From rivals to co-authors: how Booch, Rumbaugh and Jacobson invented UML (and what's left of it today)
The story of the Three Amigos of UML and RUP: how Booch, Rumbaugh and Jacobson, from rivals, unified object-oriented modeling. And what's left of it today.
Technical Leadership & Projects
5 rules I've seen work in project teams that hold
Project management: 5 rules observed in teams that hold under pressure. Psychological safety, bus factor, outcome vs output, knowledge transfer.
Technical Leadership & Projects
AI Manager and Project Management: when artificial intelligence enters the project
AI Manager: the professional role that governs the impact of artificial intelligence on architectures, processes and people. Reflections from 30 years in IT.
Oracle
Oracle
The Saturday night nobody wants: migrating 12 TB from Oracle 12.2 to 21c in a four-hour window
Moving 12 TB from Oracle 12.2 to 21c with transportable tablespaces and RMAN incremental backups, keeping downtime under four hours. Real numbers and commands.
Oracle
Research on millions of legal documents: how Oracle Text changed response times
How three Oracle Text indexes cut legal archive search times from 90 seconds to under 700ms. A real case with CONTEXT, CATSEARCH, and CTXXPATH.
Oracle
Three triggers, a nightly job, and 1,247 orphaned policies: Oracle 26ai Assertions explained
Oracle 26ai introduces CREATE ASSERTION, the cross-table declarative constraint SQL-92 promised for decades. Syntax, comparison with triggers, and real insurance use cases.
Oracle
ORA-00205 at three in the morning: understanding Oracle critical files before you need to
Step-by-step diagnosis of ORA-00205 during a night emergency: MOUNT phase, control file, SPFILE, redo log. A practical method for on-call Oracle DBAs.
Oracle
Oracle 19c, 21c, 23ai, 26ai: the silent rewrite of value domains
Seven years of Oracle seen through enumerations: from 19c CHECK to 23ai SQL Domains, all the way to 26ai Assertions. An insurance migration story.
Oracle
Enumerations in Oracle: twenty years of workarounds, and the road that opened up with 23ai
Oracle never had a native ENUM. CHECK constraints, lookup tables and SQL Domains 23ai: three roads, a real banking case, and what's coming with 26ai.
Oracle
Oracle from On-Premises to Cloud: Strategy, Planning and Cutover
Oracle 19c on-premises to OCI migration: 2 TB with RAC and Data Guard. BYOL licensing, Data Pump, overnight cutover — real chronicle.
PostgreSQL
PostgreSQL
Logical replication in PostgreSQL: a colleague's questions that clarify the topic
PostgreSQL logical replication explained through a colleague's questions: cross-version migration, CDC to data warehouse, step-by-step configuration.
PostgreSQL
ENUM in PostgreSQL: when the choice pays off, and when it bites back
PostgreSQL ENUM vs CHECK vs lookup table: ALTER TYPE ADD VALUE is metadata-cheap, dropping a value costs a migration. Three paths, one real telco case.
PostgreSQL
When an index does more harm than good: cleaning PostgreSQL waste
PostgreSQL: how to find and drop unused indexes with pg_stat_user_indexes. Real case: a table with 15 indexes, 8 of them never used.
PostgreSQL
pg_stat_statements: the first thing to install on any PostgreSQL
PostgreSQL pg_stat_statements: the first query diagnostics extension to install. Find the three queries consuming 80% of database resources.
PostgreSQL
VACUUM and autovacuum: why PostgreSQL needs someone to clean up
PostgreSQL VACUUM and autovacuum: bloat diagnosis on a 200 GB database, reading pg_stat_user_tables and tuning without disabling anything.
PostgreSQL
Roles and Users in PostgreSQL: Why Everything Is (Only) a ROLE
PostgreSQL ROLE: users and roles are the same object. Mental model, GRANT, NOINHERIT and how to build a truly maintainable read-only account.
PostgreSQL
When a LIKE '%value%' Slows Everything Down: A Real PostgreSQL Optimization Case
PostgreSQL optimization: LIKE '%value%' causes full scan. Using pg_trgm and GIN index to turn a wildcard search into a fast lookup. Real case.
MySQL
MySQL
128 MB on a 32 GB machine: InnoDB buffer pool sizing in production
A MySQL 8.0 e-commerce with 32 GB RAM running on 128 MB buffer pool. How we diagnosed the issue, fixed it, and cut query latency from 45ms to 3ms.
MySQL
The 8:40 AM phone call: recovering a Galera Cluster when nodes fall one after another
A Galera 3-node cluster loses quorum during a live incident. Diagnosis, emergency bootstrap, and recovery procedure guided over the phone — with the runbook.
MySQL
The Monday morning report: four hours of MySQL replica lag
Four hours of lag on a MySQL replica in a logistics system. How we diagnosed it, why Seconds_Behind_Master lies, and what actually fixed it.
MySQL
The ticket said 'update MySQL and shut down the service'
A real MySQL 8.0.34→8.0.45 patching workflow: broken views blocking mysqldump, GTID warnings, replica role check, and RPM upgrade order on RHEL 8.
MySQL
Swap at 100% on InnoDB Cluster: when join_buffer_size multiplies the problem
A MySQL InnoDB Cluster grinding to a halt due to join_buffer_size set to 2 GB per thread. Diagnosis, rolling restart, and the math behind per-thread buffers.
MySQL
ENUM in MySQL: when it makes your day, and when it ruins your week
MySQL ENUM vs CHECK constraint vs lookup table: three ways to model an enumeration. Pros, cons and real case of a shipment tracking system.
MySQL
Before upgrading MySQL: the numbers the customer actually asks for — and how to find them
MySQL 8.0 pre-upgrade assessment: measuring sizes, growth, backup and restore times with information_schema. Real numbers to plan maintenance window.