Backups, data transfer & maintenance

1. Define recovery goals

Recovery point objective describes tolerable data loss measured in time. Recovery time objective describes the target time to restore service. Backups, dependencies, people and procedures all affect whether those objectives are attainable.

What to expect

A one-page plan naming the service, its dependencies, acceptable loss, restore time target, backup locations, recovery owner and validation steps.

Not by itself. You need a mechanism and tested process capable of recovering a more recent point, such as appropriate transaction-log backups or another supported design.

Watch for: Do not confuse backup frequency with proven recoverability or replication with independent protection.

2. Validate an import before writing

CSV is text, not a database schema. Parse its structure, check required fields and unique identifiers, then stage validated rows before changing the destination. This fixture deliberately contains only two known contacts.

import csv, io
source = io.StringIO("id,name\n1,Ada\n2,Lin\n")
rows = list(csv.DictReader(source))
seen = set()
for row in rows:
    if not row["id"] or not row["name"] or row["id"] in seen:
        raise ValueError("Missing field or duplicate ID")
    seen.add(row["id"])
print(f"Validated {len(rows)} rows")

What to expect

Validated 2 rows. The program does not write to a database.

The validation raises ValueError. Reject or quarantine the batch and report the offending row; do not silently overwrite a different record.

Watch for: CRM is a category of application, not a universal schema or vendor. Map fields, encoding and relationships explicitly for the actual system.

3. A change needs an exit condition

A patch procedure should define its target versions, prerequisites, test fixture, observation window, expected service checks, and rollback decision. Rebooting successfully is only one check.

What to expect

A maintenance record with before/after versions, application check results, backup verification and a decision to accept or roll back.

No. Evaluate the documented service checks and dependency errors. Follow the tested rollback or repair plan and record the outcome.

Watch for: Do not assume a VM checkpoint covers an external database, directory or storage dependency.