v0.4.0

ynf v0.4.0

Each run's telemetry budget now reaches ynr, and ynf moves to the current ynh, ynr and ynm releases. Nothing has been removed, and no existing lane changes behaviour.

The spool manifest carries the run's budget: run_quota_bytes

ynf holds each run's spool folder to telemetry.run_quota (64 MiB by default). Until now that budget was never handed to ynr. The manifest ynf writes for each run (manifests/<run id>.json) now carries it as run_quota_bytes, in bytes, beside the run, lane, harness, focus, item, step and uid.

From ynr 0.3.1, ynr serve reads it: it stops reading a run folder that exceeds its budget, and stamps what it shipped from that run with ynr.provenance.warning=over-budget. A ynr that predates the field ignores it, and a manifest without it is read with the 64 MiB default.

Pins

The factory image and CI are tested against:

• ynh 0.14.0 (was 0.12.0)
• ynm 0.4.2 (was 0.4.1)
• ynr 0.3.2 (was 0.2.2)

ghcr.io/eyelock/ynf-factory:0.4.0 carries ynh 0.14.0 and ynm 0.4.2. What the new ynh brings to a factory run includes: sensors from an included harness run against that harness (and an include can opt out with "sensors": false), a harness's includes baked into its image, and the harness's own repository refreshed on every run. ynh 0.14.0 still lists the agent-run-harness-path and agent-run-fetches-includes features, so folder harnesses run as before.

ynf's CI now also runs ynr's conformance rule 13 (never into memory), with ynm on the PATH, and it passes.

Upgrading from 0.3.0

Nothing to change. run_quota_bytes is written for you from telemetry.run_quota. To have ynr enforce it, run ynr 0.3.1 or later.

Install

• brew install eyelock/tap/ynf
• ghcr.io/eyelock/ynf-factory:0.4.0