nice tests. this looks good to me, tbh! I prefer my pools to drain when my app is idle...feels better to me than leaving a bunch of sockets open doing nothing for > 5 minutes, but I can definitely see in some "needs to be fast, but gets spikey traffic" this could come in handy. That initial connection to the backend is no joke. I've seen it take upwards of a second on GCS in the past.

nice tests. this looks good to me, tbh! I prefer my pools to drain when my app is idle...feels better to me than leaving a bunch of sockets open doing nothing for > 5 minutes, but I can definitely see in some "needs to be fast, but gets spikey traffic" this could come in handy. That initial connection to the backend is no joke. I've seen it take upwards of a second on GCS in the past.

Great! What would be the next steps to get this into the next release?

Thoughts on implementing or documenting:

• prefilling the pool to the minimum
• maintaining the minimum on connection errors

?

Keepalives are also sort of related to this, but I think any documentation and work needed there stands on its own.

Thoughts on implementing or documenting:

• prefilling the pool to the minimum
• maintaining the minimum on connection errors

?

I can document that this is not implemented at the moment, and merge this as a first step. Those other things yeah could come in the future.

I will release a new minor version of pg-pool on Monday morning. i don't like doing afternoon/evening releases because I always like to be around and online throughout the day to respond to any unintentional breaking changes or bugs

Hi @brianc, I'm not seeing the update in the latest 8.15.6 release. Do you have an ETA on this?

When create connections with number more than min simultaneously, they all will be removed after idleTimeoutMillis.

@htrajan @zyf0330 I'm seeing it right here am I missing something?

@brianc nope, it's all good now. @zyf0330 might have spotted an edge case where my implementation doesn't work. Unfortunately, I'm pretty tied up now and won't have time to look into it in the near future. I was able to deploy the change to my company's application and get the desired outcome of no more pg-pool.connect spikes.