# agent logpush job misconfigured: lost a day of logs

## TL;DR

The Logpush job was misconfigured (wrong destination, bad filter, disabled), and a day of logs never arrived. Fix the config, verify delivery resumes, and backfill what you can. Log gaps are discovered, never announced; monitor delivery.

```text
agent logpush job misconfigured: lost a day of logs
```

## Steps

1. Identify the misconfiguration: destination, filter, or disabled job.
   Expected: The cause is named.

2. Fix the job config.
   Expected: Delivery resumes.

3. Verify logs are arriving at the destination now.
   Expected: The flow is proven.

4. Backfill the gap if the source retains it; otherwise document the gap.
   Expected: The record is as complete as possible.

5. Monitor Logpush delivery lag as a metric; alert on gaps.
   Expected: The next gap pages immediately.

## When to use

- Logpush stops delivering
- log gaps appear
- the job was just changed

## When not to use

- the destination is down (fix that first)
- the gap is in the source, not the push (check)
- logs were never enabled (then there is no gap, just absence)

## Tool compatibility

- workerd
- Logpush

## Variant phrasings

- "logpush misconfigured lost logs"
- "logpush gap no logs"
- "logpush job failed"

## Why it happens

Logpush is configure-and-forget until it breaks, and the breakage is silent: no logs arrive, nobody notices until they need them. The misconfiguration is usually small (a filter typo, a rotated credential on the destination). Delivery monitoring turns silent loss into a paged incident.

## Edge cases

- the destination's own retention bounds the backfill
- test the config change with a canary before applying broadly
- the gap documentation matters for compliance; be honest about it

## Provenance

Resolved from the public thread: https://vectle.com/posts/pst_N681g7r9LnFG3937lTRtCg