bug: concurrent map read and map write

Expected Behavior

I ran into a weird error in a test run, just using the temporal auto-setup docker image version 1.19.1. I don't think I'm doing anything special. Just starting some basic containers.

Actual Behavior

The following logs show how temporal failed to start
https://gist.github.com/vikstrous2/7d016b5562903b723d93b6a403589620

Steps to Reproduce the Problem

Start temporal from this docker-compose file over and over again until this error triggers:

version: '3.4'
services:
  temporal-db:
    image: postgres:9.6.24-alpine@sha256:8342bcb43446694428ec6594e72e4299692854f0fc3aca090b0ab46f4c7f32a1
    restart: unless-stopped
    environment:
      POSTGRES_PASSWORD: temporal
      POSTGRES_USER: temporal
    ports:
      - 5434:5432
    healthcheck:
      interval: 1000h
      test: 'true'
  temporal:
    image: temporalio/auto-setup:1.19.1@sha256:3b582c47c354e7f9958c098f168ceb514766ab93526e9be1d772179663710d0f
    restart: unless-stopped
    depends_on:
      - temporal-db
    environment:
      - DB=postgresql
      - DB_PORT=5432
      - POSTGRES_USER=temporal
      - POSTGRES_PWD=temporal
      - POSTGRES_SEEDS=temporal-db
    ports:
      - 7233:7233
    healthcheck:
      interval: 1000h
      test: 'true'

Specifications

• Version: 1.19.1
• Platform: docker

Start by reproducing the failure with the docker-compose configuration in the issue and inspect the linked logs for the concurrent map error. No source file or test is identified, so trace the reported startup failure through the Go service initialization; done means repeated starts no longer trigger the error and coverage verifies the fix.