HydraIssues

What

hydranode pins hydrarelease v1.17.5. pkg/updater v1.20.0 (issue #539, closed 2026-08-20) fixed the Windows restart path: update --force from a separate process now restarts the scheduled task itself (schtasks /End then /Run), and a locked .backup image no longer wedges auto-update. hydrabody picked this up in v2.0.68 and it is validated in production.

hydranode is NOT broken today, so this is not urgent. Bump the dependency when hydranode next ships a release for its own reasons:

go get github.com/cederikdotcom/hydrarelease@v1.20.0 && go mod tidy

Rollout caution for the FIRST forced update after the bump

With the new updater, "hydranode update --force" via hydracluster exec will /End the hydranode task — the process that serves the exec channel. Expected behavior, verified on hydrabody:

• The exec'd command dies with hydranode, so its result comes back lost or pending. This looks like a failure but the update worked.
• The task's recurring 1-minute TimeTrigger (pkg/body/task_xml.go, IgnoreNew policy) revives hydranode on the new binary within 60 seconds. Exec is down roughly 10 to 60 seconds per node. Self-healing, no physical access needed.

Procedure: update one node at a time, wait for the node to come back online, confirm the running version, then move to the next. Do not fire the update at every node at once — exec would drop everywhere simultaneously.

Reference: hydrarelease docs/runbooks/runbook-updater-library.md, section "Service didn't restart (Windows scheduled task)".