Cron Calendar
Explain cron expressions and preview the next run times in any timezone.
OpenTurn a wall of repetitive log output into a short, ranked list of message templates.
Drop on-call snippets or app logs into the editor.
Numbers, IPs, and UUIDs collapse into patterns.
Emails and card-like numbers are redacted locally.
Share frequency tables with your incident channel.
The on-call page fired and you have 8,000 lines from the last ten minutes. Most of it is the same health check and request log repeated with different IDs. Log Pattern Miner groups those repeats so the three odd lines that explain the outage are easy to find.
Paste one event per line. Before clustering, variable parts such as timestamps, UUIDs, IP and MAC addresses, hex values, hashes, JWTs, durations and plain numbers are replaced with placeholders. Lines are then grouped with a simplified Drain-style approach: bucketed by token count, then merged when their position-by-position similarity clears a threshold you control (0.30 to 0.95, default 0.65). Each template shows its count, share of the log, dominant level, first line number and a sample. Templates under the anomaly cut-off (1 percent by default, adjustable up to 10) are listed separately as candidates worth reading.
If near-identical lines stay in separate templates, lower the threshold; if unrelated messages merge, raise it. A line with an extra token never joins a shorter one, because of the bucketing. The whole cluster table exports as CSV. For build output specifically, CI Failure Studio also understands JUnit and SARIF, and the secret masking tool is worth a pass before you paste samples into a ticket.
Clustering runs on the page's main thread and is capped at 20,000 lines. Email addresses, card-like digit runs and common phone formats are masked before templates are built. That masking is regex-based, so review exports before sharing them outside the team.
Keyboard shortcut
Answers for this tool. For site-wide help, open the FAQ hub.
The analysis panel flags it because a single message dominating the stream often means debug logging was left on or a retry loop is spinning. It is a hint, not a verdict; a busy access log can legitimately look like that.
No. Rarity is a frequency heuristic only. A once-a-day housekeeping message is rare and harmless, while a frequent template can still be a bug. Treat the anomaly list as a reading order, starting with anything tagged ERROR or FATAL.
Yes, untick the masking option when you work with synthetic data and want original values in the templates and samples. Leave it on for production logs, since the samples and CSV export would otherwise carry real emails and numbers.