ANRs

ANRs lives in the Diagnose area of the dashboard, under Issues.

At a glance#

What it does#

ANRs — Application Not Responding — are freezes: the main thread stopped servicing input long enough that the OS noticed. This page reports them over a selectable time window with a KPI strip, a trend line, a sortable by screen table, and a list of recent occurrences. Selecting a screen opens a drawer with that screen's detail.

An ANR is not a crash. The process usually survives, so nothing appears in Crashes — which is exactly why freezes go unnoticed for so long despite being one of the most damaging things a user can experience.

When to use it#

When users describe the app as "slow" or "stuck" and the crash rate is flat. The by-screen table is the fastest route from that vague report to a specific view, because freezes concentrate: one screen doing synchronous work on the main thread accounts for most of them.

Workflow#

1. 1

   Widen the window until the trend has shape

   A short window on a low-traffic app is mostly noise. Widen it until you can see whether the rate is flat, climbing, or a single spike tied to one release.

2. 2

   Sort by screen and take the top row

   Freezes cluster by screen far more reliably than crashes do. The top row is usually the whole investigation.

3. 3

   Open the screen drawer

   The drawer breaks that screen's ANRs down further, which tells you whether the freeze is one code path or the screen being generally overloaded.

4. 4

   Fix the blocking call, not the symptom

   Almost every ANR is I/O, a large decode, or a lock on the main thread. Move it off and the row disappears.

Permissions and prerequisites#

Requires an SDK version that reports ANRs, on a platform that has the concept. No dashboard configuration is needed.

Limits and edge cases#

• Every event here is a real on-device freeze. None of it is modelled or extrapolated.

• Mobile concept. A web app's equivalent signals are long tasks and interaction latency, reported under Web Vitals instead.

• A freeze that ends in process death is reported as a crash by the OS, so it will appear in Crashes rather than here.

Troubleshooting#

No ANRs at all on a busy app. Confirm the window starts after your SDK rollout. If it does, check the reporting SDK version — ANR capture is not present in every release.

The trend spikes on one day only. Compare against your release timeline. A spike bounded by two releases is a regression you have already shipped past; one that starts and does not stop is live.

Where the data comes from#

From the SDK

• Performance

Served by

• Issues

Related#

• All issues

• Errors

• Crashes

• Logs

• Bug reports

• Diagnose overview