[omattsson]

Fixes from a live test run of stack port-forward and stack exec against k8s-stack-manager on Docker Desktop.

Changes

stack port-forward: a full server tunnel limit no longer kills working tunnels

• Before: 3 handshakes refused with 429 in a row stopped the whole command, including its open tunnels. 8 parallel downloads against a server limit of 5 gave 0 of 8 completed.
• A 429 tunnel_limit now gets 3 dial retries (0.5, 1, 2 s, portForward429Delays). That gives a tunnel that drains (up to DrainTimeout) or closes time to free its server slot. The API rate limiter's 429 (no code) is not retried.
• 3 connections in a row refused with 429 stop the command only while no other connection of the command dials or tunnels (portForwarder.active). A refusal while others are active resets the counter.
• Live: 8 of 8 parallel downloads complete, and the command keeps running.

template quick-deploy exits 1 when the stack has status error

• Without a log_id the server created the stack but did not start its deploy, for example because the namespace is still terminating. The error says why and gives the next command, stackctl stack deploy <id>.
• With a log_id the deploy started and failed at once: the error points to stack logs.
• The stack is printed first in every output mode. Server fields are sanitized. client.QuickDeployTemplateLog returns the log_id.
• The server reason comes from fix(deployer): a stop during a deploy no longer leaves the stack stabilizing k8s-stack-manager#726.

stack exec refusal hints

• The hints no longer repeat the server message. Before: "The container holds a service-account token. The container holds a service-account token, so ...".
• When the body has no error text, the hint first says what the code means.

Live tests

• TestLiveWorkflow_FullLifecycle creates a stack from the instantiated definition. Before, it deployed the definition ID and got a 404.
• FullLifecycle and BulkOperations wait for each operation, so the server no longer refuses the next one.
• The cleanup waits until no operation runs, and only a 404 counts as deleted.
• waitStackStatus fails at once on every failed status.

Docs: README, CLAUDE.md, the agent guide and the command help.

Tests

• make lint and go test -race ./... pass. The 429 tests passed 5 times in a row with the race detector.
• Live against a local backend that includes fix(deployer): a stop during a deploy no longer leaves the stack stabilizing k8s-stack-manager#726: port-forward 41/41, exec 76/76, chaos 14/14, and the live suite in heavy mode (STACKCTL_LIVE_HEAVY=1) passes, including both rewritten tests.

🤖 Generated with Claude Code

[coderabbitai]

Important

• Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

Run configuration

• Source mode: Auto
• Repository: omattsson/stackctl
• Branch: fix/live-test-findings
• File: .coderabbit.yaml
• Commit: ac17212
• Layers: Repository YAML
• Review profile: CHILL
• Plan: Advanced
• Run ID: d23ca027-ff37-45df-90bb-fa461126edeb

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

Share

• X
• Mastodon
• Reddit
• LinkedIn