Recovery targets
Agreed RPO and RTO for each system, signed off by the business.
We design backup and disaster recovery around agreed recovery targets, automate it, and rehearse it, so the worst day becomes a well-practiced procedure.
$ anvexis tldr
3-2-1 copies, encrypted, monitored, and restored on a schedule.
$ anvexis facts
Stage 01
Agree how much data you can lose (RPO) and how long you can be down (RTO).
Stage 02
Databases, file stores, and configuration, encrypted and copied offsite.
Stage 03
Alert when a backup is missing, late, or suspiciously small.
Stage 04
Step-by-step recovery instructions anyone on call can follow.
Stage 05
Restore to a clean environment on a schedule and time it.
Ransomware changes the rules
Keep at least one immutable or offline copy that production credentials can't delete. Attackers know where backups live too.
Almost every company has backups. Far fewer have tested restores. The common surprises are always the same: the backup job had been failing quietly for months, the backups were on the same server that died, the files restored but the application wouldn't start, or the restore took a full day nobody had planned for.
Disaster recovery is the discipline of making sure that doesn't happen. It covers what is backed up, where copies are kept, how quickly systems can be rebuilt, who does what during an incident, and how often all of it is tested. For Odoo, that means the database, the filestore with all its attachments, configuration, and custom code, restored together and consistent.
We design recovery around two agreed targets: how much data you can afford to lose and how long you can afford to be down. Then we prove the plan works with regular drills.
| Layer | What it protects against |
|---|---|
| Point-in-time database recovery | Accidental deletes, bad imports, faulty upgrades |
| Nightly full backups | Corruption that spreads over time |
| Off-site copies in another region | Losing a data center or cloud region |
| Immutable or offline copies | Ransomware and compromised admin accounts |
| Configuration and code in Git | Rebuilding servers and applications from scratch |
Agreed RPO and RTO for each system, signed off by the business.
Database, filestore, configuration, and secrets, kept consistent with each other.
Copies in a different location and account from production.
Backups that cannot be deleted or encrypted by an attacker with admin access.
Step-by-step recovery instructions anyone on the team can follow.
Scheduled restore tests with measured times and written results.
Stage 01
List systems, data, dependencies, and current backup coverage.
Stage 02
Agree recovery targets with the business for each system.
Stage 03
Configure backups, off-site copies, immutability, and monitoring.
Stage 04
Restore to an isolated environment and time every step.
Stage 05
Set a drill calendar and keep runbooks current.
Myth: Replication is a backup.
Reality: Replication copies mistakes instantly. A deleted table on the primary is deleted on the replica too. You need point-in-time backups as well.
Myth: Cloud providers back everything up for us.
Reality: Providers protect their infrastructure. Protecting your data, including from your own mistakes, is usually your responsibility.
Myth: Testing restores is too disruptive.
Reality: Restores are tested in an isolated environment, without touching production.
The first pitfall is backing up the database but not the files. Odoo stores attachments in its filestore, and a database restored without matching files leaves broken documents everywhere.
The second is keeping backups within reach of an attacker. If the same admin account can delete production and its backups, ransomware can take both, so copies are separated and made immutable.
The third is never timing a restore. A plan that assumes a one-hour recovery is useless if the actual restore takes nine, so every drill records how long each step really took.
| Measure | Why it matters |
|---|---|
| Last successful backup | Confirms backups are actually running |
| Measured restore time | The real recovery time from drills |
| Data loss window | Gap between backups, against the agreed target |
| Drill completion | Whether scheduled tests actually happen |
Consider a company whose backups run every night to a disk on the same server. A recovery plan adds point-in-time archiving to a separate cloud account, immutable weekly copies, and a quarterly restore drill to an isolated environment.
The first drill reveals that the filestore wasn't included and the restore took far longer than assumed. Both are fixed, the runbook is updated, and the next drill meets the targets agreed with the business.
Success is confidence backed by evidence: restores tested on schedule, recovery times measured against the agreed targets, and a runbook that anyone on the team can follow at three in the morning. When something does go wrong, recovery is a routine procedure.
A long-standing US government guide recommends the 3-2-1 rule: keep three copies of important files, one primary and two backups, on two different media types, with one copy stored offsiteSource [1]. Cloud storage makes it easy to follow, as long as the offsite copy is genuinely separate from the systems it protects.
For databases, PostgreSQL's continuous archiving supports point-in-time recovery to any moment since the base backupSource [2]. The documentation is explicit that the WAL archiving procedure should be set up and tested before the first base backup, and recommends keeping several backup sets to be certain of recoverySource [2]. We schedule restore drills so recovery time is known, not guessed.
PostgreSQL HA
PostgreSQL high availability: streaming replication, read replicas, automatic failover, connection pooling, and tested backups.
OdooSupport & maintenance
Monthly Odoo support plans with written scope, response targets, security patching, monitoring, and performance care.
SecuritySecurity & trust
How we protect client systems and data: least-privilege access, encryption, secure development, secrets handling, and incident response.
Next step
Thirty minutes with an engineer, no slide deck required. You'll leave with a clearer picture, whether or not we work together.