timeline view

Log timeline: see exactly
when errors started

See the shape of the incident.

Every entry bucketed by minute and plotted per level turns a scroll-forever log file into a shape you can actually read — a flat line until it isn't.

timeline · api-gateway.log — 06:00 to 18:00
06:0014:02 · error spike18:00

zoom & pan

Zoom into the moment
that matters

Zoom in up to 10x on any part of the range and pan across it — from an hour-wide overview down to the couple of minutes a spike actually happened in, without losing the rest of the chart.

zoomed · 14:00 – 14:06
6 minutes · zoomed 10x

what it surfaces

Errors, warnings, and info
plotted separately

One-minute buckets

Entries are grouped into one-minute intervals, so a chart with an hour's worth of logs is still sixty readable bars, not a smear.

Per-level series

Errors, warnings, and info lines are counted and plotted separately, so a spike in one level doesn't get buried under normal volume in another.

Toggleable legend

Hide a level to see what's left underneath it — useful when a chatty info stream is drowning out the handful of errors you actually care about.

Zoom and pan

Zoom in up to 10x on any part of the range and pan across it, so you can go from an hour-wide overview to a single spike without losing your place.

in practice

Find the start of an incident
without scrolling

A deploy goes out at 14:02. By 14:04 the error series on the timeline is visibly taller than every bucket around it — nobody had to be watching for it. You zoom into those two minutes, hide the info level to cut the noise, and the handful of entries left make it clear this isn't the code that just shipped — it's a downstream service timing out under the redeployed instance's cold cache. No dashboard to build first, just a shape that looked wrong and a zoom to see why.

where this lives

The timeline sits above your log

It's not a separate dashboard you build — the chart is part of the file view, directly above the entries it summarises.

logier

Explorer

logs
app.log
large-log.txt
system-2200218.log

Cloud

Dropbox
OneDrive
system-2200218.log
error.log
3
Search log messages, tags, or process
Filters
2000 / 2000
415 anomalies detected11 high severity
1
error notice
− 1× +
2
Using Apache parserLayout settings
14:02:07ERRORwebhook retry exhausted
14:02:05NOTICErequest accepted POST /v1
  1. 1

    Entries bucketed by minute and stacked per level. The spike is the incident; click into it to zoom.

  2. 2

    The entry table below stays in sync — zoom the chart and the table narrows to the same window.

  3. 3

    Level filters and search apply to both the chart and the table at once.

  4. 4

    Anomaly detection flags the spike independently, so you don't have to be watching the chart to catch it.

reading the chart

Four shapes and
what they mean

The outline of a failure carries information the individual lines don't. These four come up constantly.

A vertical wall at a single minute

Something external changed at that instant — a deploy, a DNS cutover, a certificate expiry, a feature flag flip. The cause is almost never in the code that was already running; look at what happened to the environment at that timestamp.

A gradual ramp over ten or twenty minutes

Exhaustion rather than a hard break: a connection pool, a disk, a memory ceiling, a rate limit being approached instead of hit. The interesting moment is where the slope starts, not where the errors peak.

A regular sawtooth

Almost always a retry loop, and the period between teeth is your backoff interval. Worth checking whether the retries are making the underlying problem worse.

A flat line where there should be traffic

The one people miss entirely. A process that stopped logging is not a healthy process, but it produces no errors to search for — so it's invisible to a keyword search and obvious on a timeline.

large log files

Working with a file
too big to scroll

A 2 GB log with eight million lines has no useful “middle.” Scrolling is meaningless at that scale, and a text editor that tries to render the whole thing will stall before it shows you anything. The timeline is how you navigate it: instead of reading lines, you read minutes.

Every entry in the file contributes to a bucket, so the chart covers the whole range no matter how long it is. You find the bucket that looks wrong, zoom until it resolves into individual entries, and only then start reading text — at which point you're looking at a few dozen lines instead of eight million.

faq

Common questions

keep reading

From a spike to an explanation

Find the moment. Not the middle of the file.

A shape tells you where to look before you've read a single line.