Skip to content

Fix client 502/503 on deploy: start_period 60s + lb retries - #111

Merged
buzz-code merged 1 commit into
mainfrom
claude/network-data-rum-integration-c0zvc1
Jul 14, 2026
Merged

buzz-code merged 1 commit into
mainfrom
claude/network-data-rum-integration-c0zvc1

Conversation

@buzz-code

Copy link
Copy Markdown
Owner

Summary

Same fix verified working on dnd-management-nra (#65, #66), applied here:

  • client's healthcheck start_period bumped from 30s to 60s to match server — gives Caddy's independent active health check enough time to confirm the new container healthy before Swarm's start-first rolling update retires the old one.
  • caddy.reverse_proxy.lb_try_duration: 5s / lb_retries: 3 added to client's labels — a request landing on the old container while it drains after SIGQUIT (still "running" per Docker, but refusing new connections) now transparently retries against the healthy upstream instead of surfacing 502/503 to the browser.

Test plan

  • Deploy and watch for 502/503 during a rolling update of client.

Generated by Claude Code

Same fix as dnd-management-nra#65/#66:
- client's healthcheck start_period bumped from 30s to 60s to match
  server, giving Caddy's independent active health check enough time
  to confirm the new container before Swarm's start-first update
  retires the old one.
- caddy.reverse_proxy.lb_try_duration/lb_retries added so a request
  landing on the old container while it drains after SIGQUIT (still
  "running" in Docker's eyes, but refusing new connections) transparently
  retries against the healthy upstream instead of surfacing 502/503.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
@buzz-code
buzz-code merged commit cb59026 into main Jul 14, 2026
3 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants