V2 Pool Final Fix Summary

Root Cause Analysis

The test failures are caused by:

1. Insufficient Pool Size: Tests are trying to check out 5 workers concurrently but pool only has 4 workers
2. Worker Initialization Delays: Python process startup takes ~1.5 seconds, causing checkout timeouts
3. Message Queue Pollution: Workers receive checkout requests during initialization

Critical Fixes Needed

1. Increase Pool Size for Concurrent Tests

The concurrent test spawns 5 processes but the pool only has 4 workers:

# Current setup
pool_config = [
  pool_size: 4,  # TOO SMALL for 5 concurrent checkouts
  overflow: 2,
  name: genserver_name
]

# Fix:
pool_config = [
  pool_size: 6,  # Enough for 5 concurrent + 1 spare
  overflow: 2,
  name: genserver_name
]

2. Pre-warm Workers

Initialize all workers before running concurrent tests:

# After starting pool
Process.sleep(500)  # Current delay

# Better: Wait for workers to initialize
for i <- 1..pool_size do
  SessionPoolV2.execute_anonymous(:ping, %{warm: true}, pool_name: pool_name)
end

3. Fix Worker Death Test

The worker death test is using a command that doesn’t exist (force_exit). Need to use a real command that causes Python to exit.

Implementation Status

✅ Made tests synchronous (async: false) ✅ Use start_supervised! for proper cleanup ✅ Added Process.alive? guards ✅ Increased timeouts in config ✅ Fixed process lifetime in concurrent tests ❌ Need to increase pool size for concurrent tests ❌ Need to pre-warm workers ❌ Need to fix worker death test command