Error Rate Calculator

“How fast were we failing?” Error Rate Calculator buckets errors by time and reports the rate — errors per minute — plus the peak window and a per-bucket breakdown.

Normalized rates make a 5-minute blip comparable to an all-day creep, which is exactly what on-call triage needs before touching a single error. A line counts as an error when its level is ERROR/FATAL/CRITICAL, or when the text itself contains a keyword like “timeout”, “exception” or “denied” — so unstructured logs without level tags still get picked up.

The overall rate is computed as total errors divided by the time span between the first and last parseable timestamp in your paste, in minutes — not by wall-clock time you supply. Paste exactly the window you care about, and the rate reflects that window precisely.

• Time buckets — per minute, 5 minutes, hour or day, chosen from a dropdown.
• Normalized rate — errors per minute, computed from the actual span of timestamps in your paste.
• Peak window — the single bucket with the most errors, called out by name and count.
• Error recognition — matches log levels (ERROR/FATAL/CRITICAL) and free-text keywords (error, exception, failed, timeout, denied, panic) so unlabeled logs still count.
• Per-bucket report — every non-empty bucket listed with its error count, so spikes are visible at a glance.
• Unparsed tracking — lines without a recognizable timestamp are counted separately and excluded from the rate.

1. Paste the log window you want to measure — the timestamp at the start of each line drives everything.
2. Pick the bucket interval that matches your investigation: minute-level for a live incident, hourly or daily for a trend review.
3. Calculate. The tool computes total errors, the normalized rate per minute, the peak bucket, and lists every bucket with its count.
4. Read the peak bucket to jump straight to the worst moment in the window.
5. Compare two separate pastes (e.g. before/after a deploy) by running the tool twice and comparing the rate and peak figures.

Example 1 — Blip or trend? Twelve errors packed into one minute at 14:03 is an incident, not a slow creep; the peak bucket highlights it instantly instead of averaging it away over the whole log.

Example 2 — Deploy comparison. Compute the error rate for the hour before a release and the hour after, and compare the two numbers directly since both are normalized to errors per minute.

Example 3 — SLO reporting. Copy the per-bucket counts into an error-budget chart or incident postmortem so reviewers can see exactly when the failure rate rose and fell.

Example 4 — Noisy vs urgent. A steady trickle of 1 error every 10 minutes produces a low overall rate; a burst of 20 errors in 2 minutes produces a much higher rate even with the same total count — the calculator tells them apart.

• Comparable rates — errors-per-minute normalization means a 2-minute window and an hour-long window can be compared directly.
• Spike detection — the peak bucket surfaces the worst moment without manual scanning.
• Works on unstructured logs — keyword-based error recognition catches failures even without a formal log level.
• Zero config — paste and go; no schema or format setup required.
• Private — all calculation happens in your browser.