Skip to content
EncryptTechBD
Home Industries Our Work Blog Pricing About Contact Discuss Your Project
বাংলা (BN)
Website & Infrastructure 6 min read

Website backup retention problems and how to fix them

Diagnose failing restores, vanishing retention, missing offsite copies and slow backup windows, with fixes and prevention steps.

E
EncryptTechBD · 0 comments
Cisco network switches with colourful patch cables in a rack
Photo: Jemimus / Wikimedia Commons

A backup problem usually surfaces at the worst moment. The site is broken, the client is waiting, and the backup panel shows a green tick. That green tick means a job completed. It does not mean the archive is complete, the database and files belong to the same moment, or the encryption key is available. Retention problems are mostly discovered during a restore, which is the worst time to start investigating them.

This guide is for the owner or administrator of a business website. Each section names a symptom, likely cause, immediate fix and the change that prevents a repeat. The companion twelve-point backup retention checklist covers how to set retention.

First, run a thirty-minute diagnostic

Do not change retention until you know what is stored. Record five facts: the last successful backup timestamp, the oldest restore point available, repository size against expected size, the last successful restore test, and whether the backup account is separate from the production administrator account. Then check whether the database and uploads are in the same backup set.

Record the answers in one incident note. A storage problem, a credential problem and a retention problem can all present as “the backup is missing”, and changing three settings at once destroys the evidence.

Symptom: the job is green but the restore fails

Likely cause: the archive is incomplete or inconsistent. A plugin may exclude large upload directories, a database dump may have been truncated by a timeout, or the encryption key may live only in the production application. A successful upload is not a successful backup.

Fix: restore to a staging environment, import the database, verify table counts and open a sample of media files. If the archive is encrypted, confirm that the recovery key is available outside the production server before you delete anything. If the dump is truncated, increase the timeout and write the dump to a temporary file before compression.

Prevention: schedule a restore test every quarter, test the oldest and newest restore points, and store the result with the backup record. Add a checksum or archive-size check to the job so a silent failure becomes visible.

Symptom: old backups disappear sooner than expected

Likely cause: the policy counts files rather than days, the hosting provider enforces its own limit, a lifecycle rule deletes objects after a fixed period, or deduplication has broken the chain that older restore points depend on.

Fix: inspect the repository policy and the provider's retention limit side by side. Export one older restore point to a separate location before changing anything. If the provider caps retention, move the monthly and annual copies to object storage that you control.

Prevention: write the retention policy in days, not in “number of backups”, and label each copy daily, weekly, monthly or archival. Alert on repository deletion or policy change.

Symptom: the only copy lives on the same hosting account

Likely cause: the default backup directory is inside the hosting account, and the plugin was never configured with an external destination. This is common on shared hosting because it is the path of least resistance.

Fix: copy the latest archive to a separate provider with separate credentials, then verify that the copy opens. Do not delete the original until the external restore succeeds.

Prevention: keep one copy offline or immutable and one in a different account. Use a separate backup identity that cannot administer production.

Symptom: the database and files come from different moments

Likely cause: files and database are backed up by separate jobs at different times. Orders placed between the two jobs can reference uploads that do not exist in the restored archive, or the database can point to a plugin version that was replaced later.

Fix: restore both to the same staging environment and run a reconciliation: order count, latest order number, media references and plugin versions. If they do not match, choose the older consistent pair rather than the newer mismatched pair.

Prevention: take the database dump and copy the files inside one maintenance window, then record one backup-set identifier. Busy stores need a snapshot that captures both together.

Symptom: the backup window slows the site or never finishes

Likely cause: a full backup runs during peak traffic, the archive is written to the same disk as the database, or the job has no throttle and saturates the connection.

Fix: move the job to the quietest window, enable compression and throttling, and exclude caches and logs that can be rebuilt. If the site is large, take an initial full copy and then incremental copies, with a periodic full refresh.

Prevention: size the window from the actual runtime plus a margin, and monitor job duration as a trend. A job that takes longer every week will eventually collide with the morning traffic even if it succeeds today.

Symptom: retention deletes the copy you need after an incident

Likely cause: there is no suspension mechanism for legal, regulatory or investigation holds. The lifecycle rule keeps running while the business is trying to preserve evidence.

Fix: export the relevant restore points to a separate account, record who exported them, and suspend deletion through the documented process. Involve legal counsel before deciding what to preserve.

Prevention: add a hold procedure to the retention policy, with a named approver and a review date. Test that a hold actually prevents deletion rather than merely hiding the file from the console.

Symptom: nobody can reach the backup during an outage

Likely cause: the backup console uses the same single sign-on, domain or email account as production. When that identity provider is unavailable, the recovery path is locked behind the incident.

Fix: use a separate break-glass account with credentials stored offline, and confirm that it can reach the repository without the production network. Document the sign-in path on paper.

Prevention: test the break-glass path annually and after any identity change. Keep the recovery key and the contact details for the backup provider in the offline pack.

A retention model that avoids the repeat

Separate retention into layers instead of arguing about one number. Operational copies cover the last few days; weekly copies cover a longer working horizon; monthly or archival copies cover slow corruption, legal holds and major rebuilds. Set each period from the recovery point objective and how long a bad change might go unnoticed.

For each layer, record the data included, storage location, key custodian, restore test date and deletion rule. Test the oldest copy, not only the newest, and keep one immutable copy for a period that matches your risk decision.

When to escalate

If files are being renamed, backups are disappearing or a ransom note appears, stop treating this as a retention issue. Disconnect the affected systems, preserve the logs and contact an incident responder. Do not delete “infected” backups; they may be the only clean recovery point. Our managed technology team handles backup configuration, restore testing and monitoring, and the pricing page explains how support is scoped.

Sources

Services that deliver this

Want this implemented properly?

Pricing →

Every engagement starts with a written scope. If a requirement is outside it, we quote it separately before starting.

Responses (0)

No comments yet. Add the first one below.

Leave a response

Your email address will not be published. Required fields are marked *.

Keep reading

More field notes