How we start every morning with fixed bugs from an Ona Sentry Automation
Noisy Sentry alerts cause alert fatigue and wasted spend. See how Ona automates triage, fixes, and PRs daily.
Here's how we built an Ona Automation that triages Sentry issues, fixes the underlying code, and has PRs ready to merge before we start our day.
The problem
Error monitoring tools like Sentry are only as useful as their signal-to-noise ratio, and that ratio degrades over time. Dependencies introduce harmless warnings, deprecated APIs log errors that don't affect users, and retry logic generates duplicate events. Each issue is too small to prioritize on its own, so the noise accumulates until engineers start tuning alerts out entirely.
This creates two costs. The direct cost is financial: Sentry bills by event volume, and at enterprise scale noisy issues across hundreds of services can drive thousands of dollars per month in spend with no corresponding value. The indirect cost is operational: when most alerts are noise, real incidents are easier to miss and response times increase.
Engineers know they should fix these issues. They just never do, because the work is tedious and always loses out to product backlog priorities.
The solution
We built an Ona Automation that triages our Sentry issues and fixes them automatically. The automation connects to our Sentry instance, identifies issues worth fixing, makes small targeted code changes, and opens PRs ready for review.
The core of the automation is an Ona skill, which means it's not just available as a scheduled automation but also usable interactively inside Ona environments. Our engineers use the same skill when they want to manually triage and fix Sentry issues during their day, and the automation runs it on a schedule so it also happens without anyone thinking about it.
How the automation works
The automation can be triggered manually or set to run on a schedule. We run ours every morning at 9:30am. When it fires, the skill connects to our Sentry instance, reviews open issues, identifies ones it can fix, and opens a PR with the change.
The fixes are intentionally small. A single noisy error log that's been cluttering our Sentry dashboard gets a targeted code change that resolves the underlying issue. The agent doesn't try to tackle everything at once, it picks issues it's confident it can fix cleanly, makes the change, and opens a PR.
Incremental by design
Like our other cleanup automations, this one works by chipping away at the problem daily rather than attempting a big-bang fix. Each PR is small enough to review in a couple of minutes. Engineers merge the good ones and close anything that doesn't look right. Over time, the backlog of Sentry issues shrinks to zero. We've already reached a point where the automation has fixed everything it can find.
The result
The practical outcome is simple: we start every workday with a couple of Sentry bugs already fixed and ready to merge. Our morning focus goes to product backlog rather than triaging and fixing monitoring noise.
At Ona, this has already saved us hundreds of dollars per month on our Sentry bill, and we're a small team. For enterprise organizations where noisy issues account for a significant share of event volume, the same approach translates to thousands per month in direct savings before accounting for the operational impact of cleaner alerts.
The compounding effect matters too. Noise inevitably returns as code changes, but because the automation runs daily it catches new issues before they accumulate. Our Sentry dashboard stays clean, alerts stay actionable, and on-call rotations stay sustainable.
How to set this up
The automation runs on a schedule in a secure, isolated environment with access to both your Sentry instance and your codebase. Because it's built as a skill, your team can also use it interactively when they want to fix Sentry issues on demand. Like all Ona Automations, it runs in your infrastructure so no code or Sentry data leaves your network, and it opens PRs for human review without ever merging or deploying on its own.
If you're dealing with a noisy Sentry environment, get in touch. We're actively rolling this out and would love to work with teams tackling this problem.