Audit Upstash Redis backups daily and fix gaps automatically
Every morning, check that every Upstash Redis database has a recent backup, take one where it is missing, and raise a Linear issue only when a person is needed.
Upstash
Linear
Slack BotEngineeringOperationsNotifications & AlertsResearch & MonitoringEvery morning at 6am, audit backup coverage across all of my Upstash Redis databases. Fix whatever can be fixed automatically, and escalate only the cases a human actually has to look at.
Start by calling Upstash List Redis Databases to get every database in my account. Skip any database whose name marks it as throwaway, so short lived environments do not generate noise. By default, treat a name as throwaway if it contains ephemeral, preview, tmp, temp, scratch, sandbox, pr- or -test. Keep track of what you skipped so it can be mentioned in the summary, but take no action on those.
For each remaining database, call Upstash List Backups and classify it into one of three states. Healthy means there is at least one successfully completed backup from within the last 24 hours. Stale means the most recent successful backup is older than my freshness threshold, which defaults to 24 hours. Uncovered means there are no backups at all, or none that ever completed successfully. A backup that exists but failed, or is still stuck in progress, does not count as usable.
For every stale and uncovered database, try to self heal before escalating anything. Call Upstash Create Backup to take a backup immediately. Then, if that database does not already have automatic daily backups turned on, call Upstash Enable Daily Backup so the same database does not show up as failing again tomorrow. Taking one backup fixes today; turning on daily backups fixes the pattern.
After each repair attempt, call Upstash List Backups again for that database to confirm whether a usable backup now exists. Record the outcome of every attempt, including the exact error text and status code when an attempt fails.
Escalate to Linear only when the problem survives the repair attempt. Escalate if the database still has no usable backup after the attempt, if the plan does not support backups, if the database is in a state that blocks backups, or if this database has been failing repeatedly. Before filing anything, call Linear Search Issues to look for an open issue that already names this database. If one exists, add a comment with today's findings using Add Comment to Issue instead of opening a duplicate, and treat that existing issue as evidence this is a repeat offender rather than a one off. Only call Linear Create Issue when there is no open issue for that database yet.
Whatever you write in Linear has to stand on its own as actionable work. Name the database, its region and its plan. State when the last good backup was, or state plainly that there has never been one. Then state explicitly what you already attempted and exactly what happened, including the error returned, so the assignee does not waste time repeating a backup attempt that has already failed. The issue should make it obvious what is left for a person to decide, such as upgrading the plan or retiring an abandoned database.
Finish with exactly one summary message using Slack Bot Send a Message to my infrastructure channel. Cover the healthy count, what was self healed this run and is now confirmed covered, and anything escalated to Linear with a link to the issue. Mention the databases skipped as ephemeral in a single short line. If everything was already healthy and nothing needed action, say that in one line rather than padding the message, so that the mornings it does have something to report are the ones worth reading.
Example output
What does this prompt do?
- Checks every Redis database in your Upstash account each morning and sorts them into healthy, stale, and completely uncovered.
- Takes a fresh backup on the spot for anything stale or uncovered, and switches on automatic daily backups so the same database does not fall behind again tomorrow.
- Opens a Linear issue only for what it could not fix, naming the database, its region and plan, when the last good backup was, and everything already attempted.
- Posts one Slack summary covering the healthy count, what it repaired by itself, and what it handed off to a person.
What do I need to use this?
- An Upstash account with a management key, so the workflow can see your Redis databases and their backups
- A Linear workspace, and the team where backup problems should be filed
- A Slack workspace and a channel for the morning summary
- Backups are only available on Upstash plans that support them. Databases on a plan without backup support get escalated rather than repaired.
How can I customize it?
- Freshness threshold: a backup counts as healthy if it is under 24 hours old. Tighten it to 12 hours for critical data, or loosen it to 48 hours if a daily run is more than enough.
- Which databases to ignore: by default any name containing ephemeral, preview, tmp, scratch, sandbox, pr- or -test is treated as throwaway and skipped. Edit that list to match how your team names short lived environments.
- Whether to switch on daily backups: the workflow turns on automatic daily backups when it repairs a database, so the problem stops recurring. Remove that step if you would rather approve backup settings by hand.
- Timing and destinations: the 6am run time, which Slack channel gets the summary, and the Linear team, assignee, and priority used for escalations.
FAQs
Will this work if some of my databases are on a free plan?
Will it create a duplicate Linear issue every single morning?
Does it read or change the data inside my databases?
What happens on a morning when everything is fine?
Why does it try to fix things instead of just alerting me?
Related templates
When someone leaves, we check which shared passwords they used in their final months and post a ranked rotation list to your security channel.
Every Monday, check every S3 bucket for public exposure, missing encryption and weak backup settings, then get the risks ranked in Slack.
Twice every weekday, the conversations from your social inbox land on the right HubSpot contact timelines, with a Slack recap for sales.
Every morning, find the addresses that hard bounced or filed a spam complaint, update the matching HubSpot contacts, and post a short Slack recap.
Every weekday at 4pm, spot the threads that went quiet, stage a ready-to-send nudge in your mailbox, and get a ranked Slack recap.
When you merge a fix in GitHub, this agent checks the matching dead-letter queue, replays the failed messages, and reports back on the pull request and in Slack.
Stop discovering missing backups in the middle of an incident.
Let the audit run at 6am, repair what it can, and put only the real problems on someone's board.