๐Ÿ› ๏ธ Developer ยท Updated November 20, 2026 ยท 8 min read

Epoch Timestamps in API Logs: Tokens, Retries, Skew

exp 1789000000 expired 2h ago ๐Ÿž

When an API call fails with a 401, a 429 or a rejected webhook, the answer is usually hiding in an epoch number: a token's exp claim, a rate-limit reset header or a signature timestamp. Paste that value into a timestamp converter, read the Relative line, and you know in seconds whether it is in the past, the future or just outside a tolerance window. This playbook covers the values you meet while debugging API logs, how to line up services that log in different precisions, and the sentinel numbers that look like bugs but are not.

โฑ๏ธ Try Unix Timestamp Converter now โ€” freeOpen โ†’
Result rows for 1790553600 in UTC: Monday, September 28, 2026 at 12:00:00 AM UTC with the Relative row reading 9 days ago
Read the Relative row to see in seconds whether the token is expired or still valid.
๐Ÿ’ก Why epoch values hide the root cause

Machines agree on epoch time and humans cannot read it, so the one field that explains an outage is often the one nobody looks at. A JWT that expired 40 seconds ago, a server clock that drifted two minutes, a retry that fired before the reset window, or a webhook replayed from a queue five minutes late all look identical from the outside: a request that should work and does not. The raw numbers in the log already contain the explanation, but only once you convert them and compare them with now and with each other. Doing that inside a browser tab also matters on a production incident, because token payloads, customer IDs and internal hostnames in the lines you copy should not be pasted into a random website that logs requests.

Token claims: exp, iat and nbf

JSON Web Tokens store times as NumericDate values, which are plain seconds since 1970. Decode the token first with a JWT decoder, then paste each claim into the converter and read the Relative row.

A quick check for skew: copy now from the converter's live clock, pause it, and compare with the Date header of a response from the failing server. A gap of more than a few seconds means the host needs NTP fixed, and no amount of token debugging will help.

Rate limits, retries and webhook signatures

Throttling and signature checks are where precision and units cause the most confusion, because providers disagree on formats.

When a signature check fails only in production, convert the signed timestamp and the time your handler ran. If the gap is minutes, the payload waited in a queue or a retry, and the fix is to verify on receipt rather than after processing.

Debugging API logs written in mixed precisions

A single request can cross services that each log time differently: a Node gateway in milliseconds, a Python worker in microseconds, a Go service or tracing backend in nanoseconds. The converter detects the unit from the digit count, so you can paste each raw value without trimming.

Sub-millisecond detail is not kept, so two events logged 300 microseconds apart will show the same ISO value. For ordering at that resolution, compare the raw integers directly after padding them to the same unit.

When logs include noisy lines, the Regex Tester can pull out the timestamp fields before you convert them.

Sentinel values that look like bugs

Some numbers show up constantly in logs and have nothing to do with real events. Recognizing them saves a lot of time.

If a cache entry, session or feature flag claims an expiry from one of these values, the bug is upstream in whatever wrote the field, not in the clock.

Step-by-step

1234
1Copy the raw number from the log line, header or decoded token claim, without quotes or units.
Token exp value 1790553600 pasted into the timestamp converter with Detected: Unix seconds under the box
Copy the raw exp number from the log or decoded token, with no quotes or units.
2Paste it into the converter, confirm the detected unit, and set the zone to UTC to match your servers.
UTC quick zone selected so the converted time matches server clocks, with date and ISO rows updated
Set the zone to UTC to match your servers.
3Read the Relative row to see how far the value is from now, and compare with the Date header or your log's own time.
Result rows for 1790553600 in UTC: Monday, September 28, 2026 at 12:00:00 AM UTC with the Relative row reading 9 days ago
Read the Relative row to see in seconds whether the token is expired or still valid.
4For many lines, use Batch mode, download the CSV and sort by the ISO UTC column to rebuild the incident timeline.
Batch convert panel with four log epoch values 1790553600 to 1790553781 converted to ISO 8601 UTC times one minute apart
For many lines, use Batch mode and sort by the ISO UTC column to rebuild the timeline.

Common mistakes to avoid

โš ๏ธTreating expires_in or a rate-limit countdown as an epoch value and concluding the token expired in 1970.
โš ๏ธBlaming a token when nbf or iat is slightly in the future; that is clock skew on one of the hosts.
โš ๏ธSorting a mixed-precision export on the Unix seconds column, which collapses events inside the same second.
โš ๏ธPasting production tokens and logs into converters that send your input to a server.

Pro tips

โœ“Keep your incident timeline in UTC with ISO strings; switch to a local zone only for the customer-facing summary.
โœ“Pause the live clock when you capture now, so the reference point in your notes does not keep moving.
โœ“Log the unit in the field name, such as ts_ms or created_ns, so the next on-call engineer does not need to guess.
โœ“When two services disagree by exactly 3,600 seconds, suspect a local time written as if it were UTC.
โœ“Use the Date and time panel to build a test token exp exactly 5 minutes ahead in UTC for reproducing expiry bugs.

Frequently asked questions

Are JWT exp and iat in seconds or milliseconds?

Seconds. The JWT standard defines them as NumericDate, whole seconds since 1970 UTC. A 13-digit exp usually means a library wrote milliseconds by mistake.

Can the converter read HTTP date headers?

Yes. Paste values like Mon, 28 Sep 2026 14:30:00 GMT from Retry-After, Expires or Last-Modified and it converts them to epoch and ISO forms.

How do I compare nanosecond and millisecond log lines?

Paste both; the tool detects each unit by digit count and shows them on the same UTC timeline, down to the millisecond.

Why does 253402300799 show 1978 in the main results?

Twelve digits are read as milliseconds by default. Because that length is ambiguous, the detected line also shows the seconds reading, 9999-12-31T23:59:59.000Z, the usual never-expires sentinel.

Do my tokens or logs get uploaded?

No. Conversion uses your browser's own date functions, so nothing you paste leaves the tab.

๐Ÿ“Œ Bottom line

Most 401s, 429s and rejected webhooks come down to one epoch number compared with the wrong now. Convert the claim or header, read how far it is from the present, check the host clocks for skew, and use Batch mode with UTC and the ISO column when you need a timeline across services that log in different precisions.

Open Unix Timestamp Converter โ†’

Related guides

Browse more: all text and developer guides ยท Unix Timestamp Converter