Description
The postUpgradeReconcileWindow constant is hardcoded at 90 seconds. The hint text emitted on a version mismatch even acknowledges this is too short for slow setups: "the node is slower than the 90s reconcile window — large image pulls or cold hardware can exceed it. Re-run talm get version after a minute to distinguish: if the version updated, the node was just slow".
An operator with known-slow hardware (cold OCI instances, large extensions: set, slow registry) should be able to widen the window for their environment without forking talm or running the manual check on every upgrade.
Scope: add --post-upgrade-reconcile-window <duration> to talm upgrade (default 90s for backward compatibility). Thread it through runPostUpgradeVersionVerifyInner, which already takes the window as a parameter. Reflect the active value in both the "waiting Xs for the node to finish booting..." line and the two-hypothesis hint so a custom window is self-documenting in the output.
Activity
fix(upgrade): fail-fast on bad reconcile-window before upgrade RPC (#190
fix(apply): per-node prefix on insecure path via GlobalArgs.Nodes fal…
fix(apply): Wireguard secret paths + apply-safety-gates test plan rows (
docs(readme): align apply-safety-gates bullets with #189/#190/#192 ch…
fix(apply): address branch-review LGTM-2 blockers (#189+#190)
feat(apply,upgrade): apply-safety follow-ups (#189 + #190 + #191 + #192)
feat(apply,upgrade): apply-safety follow-ups (#189 + #190 + #191 + #192)
feat(apply,upgrade): apply-safety follow-ups (#189 + #190 + #191 + #192)
Merge pull request #200 from cozystack/feat/apply-safety-followups