3.x: PoolingOptions.maxQueueSize is too small

[dkropachev]

Currently it is 256, which means that driver allows only 256 pending requests.
Given the fact that there are nodes in the wild with 128 shards it is too low.
We need to fix it, probably it makes sense to configure it based on number of shards, or connections in the pool

[mykaul]

Currently it is 256, which means that driver allows only 256 pending requests. Given the fact that there are nodes in the wild with 128 shards it is too low. We need to fix it, probably it makes sense to configure it based on number of shards, or connections in the pool

We have a node in the 'wild' with 239 shards as well. More than I expected actually.

[Bouncheck]

To be clear this is the size of the waiting queue. This is how many queries can wait at a point in time for a connection to a host that can handle them. Otherwise the next host will be tried. The connections have a separate limit for the number of queries each of them can handle concurrently.
Increasing the queue should not affect the throughput.
The higher the amount of shards the less needed this queue should be (assuming all other variables are constant and assuming minimum of 1 connection per shard).

[Bouncheck]

I'm not sure what the ideal number would be, depends on the client's memory. I feel like there is no one size fits all, hence it should be configured by the user.

[dkropachev]

I'm not sure what the ideal number would be, depends on the client's memory. I feel like there is no one size fits all, hence it should be configured by the user.

I would say 256 per shard.