Problem
run_rollback's missing-target error (edgezero-cli/src/lib.rs:517-520 at v0.0.6) tells the operator:
a production rollback requires --rollback-to (the version to re-activate). Fastly exposes no metadata to infer it, so it must be captured before the deploy that superseded it -- use deploy's previous-version output.
No previous-version output exists anywhere in the tree: a production deploy emits only the activated version (log::info!("version={version}"), edgezero-adapter-fastly/src/cli.rs:4811). An operator who follows the hint at the worst possible moment (mid-incident, needing a rollback) finds nothing to use.
The deploy GitHub action already works around this by running active-version before deploying to capture the rollback target.
Proposal
Either:
- Emit
previous-version=<N> from a production deploy (capture active-version before activation), making the hint true and giving scripts a machine-readable line to capture; or
- Reword the hint to match reality: capture the target with
active-version before deploying.
Option 1 is strictly better for scripted deploys; option 2 is the one-line fix.
Problem
run_rollback's missing-target error (edgezero-cli/src/lib.rs:517-520at v0.0.6) tells the operator:No
previous-versionoutput exists anywhere in the tree: a production deploy emits only the activated version (log::info!("version={version}"),edgezero-adapter-fastly/src/cli.rs:4811). An operator who follows the hint at the worst possible moment (mid-incident, needing a rollback) finds nothing to use.The deploy GitHub action already works around this by running
active-versionbefore deploying to capture the rollback target.Proposal
Either:
previous-version=<N>from a production deploy (captureactive-versionbefore activation), making the hint true and giving scripts a machine-readable line to capture; oractive-versionbefore deploying.Option 1 is strictly better for scripted deploys; option 2 is the one-line fix.