-
Notifications
You must be signed in to change notification settings - Fork 26
talm reset default --wipe-mode=all destroys META; selective wipe is the friendlier default #185
Copy link
Copy link
Closed
Labels
area/commandsIssues or PRs related to pkg/commands (CLI subcommands, flag parsing, root detection)Issues or PRs related to pkg/commands (CLI subcommands, flag parsing, root detection)area/docsDocumentation / README / inline help / hint copyDocumentation / README / inline help / hint copykind/documentationCategorizes issue or PR as related to documentation, README, hint copy, error-message UXCategorizes issue or PR as related to documentation, README, hint copy, error-message UXpriority/important-soonMust be staffed and worked on either currently, or very soon, ideally in time for the next releaseMust be staffed and worked on either currently, or very soon, ideally in time for the next releasetriage/acceptedIndicates an issue is ready to be actively worked onIndicates an issue is ready to be actively worked on
Metadata
Metadata
Assignees
Labels
area/commandsIssues or PRs related to pkg/commands (CLI subcommands, flag parsing, root detection)Issues or PRs related to pkg/commands (CLI subcommands, flag parsing, root detection)area/docsDocumentation / README / inline help / hint copyDocumentation / README / inline help / hint copykind/documentationCategorizes issue or PR as related to documentation, README, hint copy, error-message UXCategorizes issue or PR as related to documentation, README, hint copy, error-message UXpriority/important-soonMust be staffed and worked on either currently, or very soon, ideally in time for the next releaseMust be staffed and worked on either currently, or very soon, ideally in time for the next releasetriage/acceptedIndicates an issue is ready to be actively worked onIndicates an issue is ready to be actively worked on
Problem
talm resetwithout explicit--wipe-mode/--system-labels-to-wiperuns with the upstream default--wipe-mode=all, which wipes the META partition along with STATE and EPHEMERAL. With META gone the node loses its bootstrap config + machine-config persistence, and cannot self-recover on the next boot — it comes up in maintenance mode and needs a full re-apply.Selective wipe via
--system-labels-to-wipe=STATE,EPHEMERALleaves META intact; the node re-joins automatically on reboot.Neither path is wrong — but the default is the destructive one, and the README/help text doesn't make the distinction clear.
Reproduction
Verified on dev17 node2: selective wipe → node rebooted, came back as a fresh etcd member from META within ~90s, with the placeholder hostname
talos-<id>until the next apply refreshed it.Expected
Either:
--wipe-modeand--system-labels-to-wipeto make the selective-vs-all trade-off explicit, plus a one-paragraph README note in the reset section.STATE,EPHEMERAL) so the friendly path is the default. Operators who actually want full wipe pass--wipe-mode=all. Riskier change — upstream defaults differ — but the safer default matches operator expectations.Option 1 is the minimum; Option 2 is the safer one if cozystack is willing to diverge from upstream defaults here.
Why this matters
The pattern "I reset the node, why doesn't it come back?" is one of the most common talm support questions. Selective wipe is the operator-friendly default that lets the node self-heal. The current default surprises operators with a full wipe.
Surfaced during the dev17 manual test plan exercise (section H).