Database Migration: Strategies, Risks, and Guarantees
Database migration is technically the riskiest stage of infrastructure work. Data loss or site unavailability during the transfer has direct financial consequences: one hour of downtime for an online store can translate into tens of thousands of dollars in lost revenue, and damaged order tables can be an irreparable blow to reputation. Experienced engineers know: 80% of problems arise from unaccounted dependencies, DBMS version incompatibility, or lack of a test rollback. Our team performs turnkey migrations with integrity guarantees and minimal downtime, backed by over a decade of practice and dozens of successful projects. We don't just copy data — we design a safe transition with zero business risk.
How to choose a migration strategy based on downtime?
The choice of strategy is driven by database size, acceptable downtime, and consistency requirements. Below is a comparison of three main approaches.
| Strategy | Downtime | Complexity | Data Loss Risk |
|---|---|---|---|
| Maintenance window (dump + transfer) | Hours | Low | Medium (manual dump) |
| Online migration (replication) | Seconds | High | Low (automatic sync) |
| Blue-Green (parallel database) | Zero | Very high | Very low (dual write) |
Maintenance window is the simplest method: put the site in maintenance mode, dump, transfer to the new server, then launch. Suitable for databases up to 10 GB during nighttime. Online migration uses replication: for MySQL — Percona XtraBackup or binlog replication; for PostgreSQL — pglogical or pg_basebackup + WAL shipping. Blue-Green requires two parallel databases — the application writes to both, then switches instantly.
What's included in the database migration work?
- Audit of the current database: schema, size, dependencies, read/write speed. We analyze index types, triggers, and stored procedures.
- Development of a migration plan with strategy selection, test run in an isolated environment, and rollback scenario.
- Replication setup (if applicable) with lag monitoring.
- Execution of the transfer during the maintenance window with minimal user impact.
- Integrity validation — comparing row counts, checksums, and application testing.
- Documentation and training — transferring schema, configs, and maintenance instructions.
We develop a detailed plan for each project, including a test run on a data copy. This minimizes risks.
Why online migration cuts downtime by 10x?
With maintenance window, downtime consists of dumping, transferring, and restoring. For a 50 GB database, this can take 4–5 hours. Online migration with pglogical reduces downtime to seconds: you just switch the application to the new server after full synchronization. Time savings reach 90%. Additionally, online migration continuously syncs changes until the switch, eliminating data loss. This method is especially effective for high-load projects where every hour of downtime is costly.
How is post-migration validation performed?
After the transfer, we automatically compare critical tables:
for table in users posts orders products; do src=$(mysql -h source -u root -p -se "SELECT COUNT(*) FROM mysite.$table") dst=$(mysql -h target -u root -p -se "SELECT COUNT(*) FROM mysite.$table") if [ "$src" != "$dst" ]; then echo "MISMATCH: $table: $src vs $dst" else echo "OK: $table: $src rows" fi done We also check checksums for text fields and test application functionality for 24 hours. If a discrepancy is found, we immediately roll back using a pre-prepared scenario. Data integrity guarantee is contractually binding.
Estimated timelines
| Database Size | Maintenance window | Online migration |
|---|---|---|
| Up to 1 GB | 2–10 min | 1–5 days (planning + execution) |
| 1–10 GB | 10–60 min | 1–5 days |
| 10–100 GB | 1–8 hours | 1–5 days |
| 100 GB+ | Not recommended | 1–5 days |
Timelines include preparation, test run, and backup. Cost is calculated individually after project assessment.
Common migration mistakes and how to avoid them
- DBMS version incompatibility: check MySQL/PostgreSQL versions on source and target. Different major versions can break indexes.
- Loss of triggers and procedures: use
--routines --triggersflags in mysqldump or--no-ownerin pg_dump. - Lack of rollback plan: always take a full dump before starting and keep it for 48 hours after migration.
Example command for MySQL
# Dump with locking for consistency mysqldump \ --single-transaction \ --routines \ --triggers \ --events \ --hex-blob \ --default-character-set=utf8mb4 \ -u root -p mysite_db \ | gzip > /backup/mysite_$(date +%Y%m%d_%H%M%S).sql.gz Example command for PostgreSQL
# Custom format (faster, compressed, parallel restore) pg_dump \ -U postgres \ -d mysite_db \ -F custom \ -f /backup/mysite_$(date +%Y%m%d).dump \ --verbose According to PostgreSQL recommendations, logical replication (pglogical) ensures minimal lag. Assess your project — contact us for a free consultation. We'll prepare a migration plan, perform a test run, and ensure a seamless transition. Order a turnkey migration: get a quote within 24 hours.







