# Troubleshooting Monitoring

Fixes for empty charts, alarms that stay in Insufficient data, webhooks that fail, empty logs and checks that are down or regional.

## A chart says "No datapoints in this range yet."

**Cause:** The machine is stopped, was started in the last few minutes, or runs in a Hetzner
project of your own whose token no longer works. Only running machines are collected.

**Fix:** Start the machine, wait a few minutes, or renew the project's token.

## An alarm stays in Insufficient data

**Cause:** The metric has no point in any of the evaluated periods: the machine is stopped, the
series is not published, or the period is shorter than the interval at which the metric arrives
(`disk_used_percent` every few minutes, `connections` every two minutes).

**Fix:** Use a period of 15 minutes or more for the slower metrics. For a custom series, check that
the alarm's namespace, metric name and dimensions match the series exactly (see
[Alarm on a custom metric](alarm-on-a-custom-metric.md)).

## An alarm on a custom series never goes to ALARM when the series stops

**Cause:** A missing datapoint never crosses a threshold.

**Fix:** Publish a zero while there is nothing to count, and alarm on `sum < 1`.

## The history says "webhook failed: refusing to follow a redirect (HTTP 301)"

**Cause:** The webhook URL answers with a redirect.

**Fix:** Use the URL it redirects to.

## The history says "webhook failed: … resolves to a private or reserved address"

**Cause:** The webhook's host name points at a private network. Webhooks reach public addresses
only.

**Fix:** Expose the receiver on a public address.

## An app's logs are empty

**Cause:** The app writes its logs to a file instead of stdout and stderr, or it has printed
nothing in the range.

**Fix:** Log to stdout and stderr, or widen the range. Lines from before the agent on the machine
started stay on the machine.

## A check is regional

**Cause:** Fewer locations than the check's `confirmations` fail: a network problem between one
location and your site, or a firewall that lets only some locations in.

**Fix:** Allow the addresses `dcloud monitoring checks locations` lists. To be told about one
failing location, set `alertOnRegional`.

## A check is down, and the site works in your browser

**Cause:** Read the reason under the check's state, or the **Detail** of its latest results: a
status outside the expected range (a `401` or `403` to requests from the platform's addresses),
a keyword that is not in the first 1 MiB of the body, a certificate that does not validate, or a
host that resolves to a private address.

**Fix:** Change the expected status, the keyword or the URL (see
[Change an uptime check](change-an-uptime-check.md)).

## A check is refused with "… is already checked … times a minute by …"

**Cause:** Checks of every organization already request the host six times a minute.

**Fix:** Choose a longer interval.

## Related

- [How alarms work](how-alarms-work.md)
- [How uptime checks work](how-uptime-checks-work.md)
