Skip to main content
Ops & Monitoring

A backup job reporting success is not a backup

Backups fail quietly in ways the job never notices. A source path changes, and the job successfully backs up an empty directory. A retention rule deletes the only copy. A job runs on a schedule that no longer matches the data it was meant to protect.

Verification asks different questions than the job: does a file exist for the period, is its size within the range of the previous ones, and is it where the restore procedure expects to find it.

None of that proves a restore will work. Only a restore proves that. It does catch the failures that otherwise get discovered during a restore, which is the worst possible moment.

The report should be short and negative: what is missing, what is the wrong size, what has not changed since last week. A list of successes is the part nobody needs.

How the ActionFlow runs on the canvas

Metadata arrives through an S3, Google Drive, Dropbox, or FTP node depending on where the backups land, listing objects rather than downloading them.

A Start Scheduled node runs the verification on your cycle, and a Loop node walks the declared backup set: the list of what should exist, never the list of what does.

A Date node computes the expected object name and period for each entry, so an absence can be detected rather than inferred.

A Validate node checks each present object against its rules: recency, size range, and location against the restore procedure.

A Math node compares size to the trailing average, because a backup suddenly a tenth of its usual size is the classic silent failure.

A Filter node keeps only the failures, and an If node marks anything entirely missing as the highest finding, regardless of its other properties.

A Generate Text node writes the report as an exceptions list, with a single line confirming how many entries passed.

A Slack node posts it, and an Email node sends the record to whoever signs off your recovery obligations.

Nodes this ActionFlow uses

  • Start Scheduled (Start): the verification cycle.
  • S3 (Storage): lists backup objects without downloading them. FTP, Google Drive, and Dropbox fit the same slot.
  • Loop (Control): walks a declared backup set, so absence is detectable.
  • Date (Util): the expected object name and period for each entry.
  • Validate (Data): recency, size range, and location against the restore procedure.
  • Math (Util): size against the trailing average, catching silent shrinkage.
  • Filter (Data) with If (Control): exceptions only, with missing ranked highest.
  • Email (Communication): the record for whoever signs off recovery obligations.

Frequently asked questions

Start building AI workflows

Create a free account, open a template or a blank canvas, and run your first ActionFlow.

Newsletter

Get product updates

New nodes, agents, and product notes. We send mail only when we have something worth opening.

Unsubscribe at any time.