PostgreSQLConnection.connect blocks for almost 6 seconds, even though it returns a Future

[sp-1234]

I tried this code (in Kotlin, but I guess it shouldn't matter), All it does: creates PostgreSQLConnection with all default values, calls connect method on it, does nothing with the returned Future, and measures elapsed time. If this call has to be blocking then maybe there's little point in returning Future because it doesn't save anything here, or maybe it's just a bug.

import com.github.mauricio.async.db.postgresql.PostgreSQLConnection
import com.github.mauricio.async.db.postgresql.column.PostgreSQLColumnDecoderRegistry
import com.github.mauricio.async.db.postgresql.column.PostgreSQLColumnEncoderRegistry
import com.github.mauricio.async.db.postgresql.util.URLParser
import com.github.mauricio.async.db.util.ExecutorServiceUtils
import com.github.mauricio.async.db.util.NettyUtils
import kotlin.system.measureTimeMillis

fun main(args: Array<String>) {
    println(measureTimeMillis {
        PostgreSQLConnection(
                URLParser.DEFAULT(),
                PostgreSQLColumnEncoderRegistry.Instance(),
                PostgreSQLColumnDecoderRegistry.Instance(),
                NettyUtils.DefaultEventLoopGroup(),
                ExecutorServiceUtils.CachedExecutionContext()
        ).connect()
    }) // prints something close to 5862 on my machine
}

I can translate it to Scala if you wish, that's not a big deal.

I'm using this library version: 'postgresql-async_2.12', version: "0.2.21"

[sp-1234]

Also I noticed that it blocks a thread for such a long time only the first time. Subsequent connections (construction of PostgreSQLConnection + connect call) only block for ~0.45ms, so it's probably OK for long running services.

Still it's interesting what exactly is happening at first connect.

[fwbrasil]

Probably related to https://github.com/mauricio/postgresql-async/issues/91

[mauricio]

Not really, connect doesn't block, only if using a pool. The actual connect operation just sets up netty and the channels: https://github.com/mauricio/postgresql-async/blob/master/postgresql-async/src/main/scala/com/github/mauricio/async/db/postgresql/codec/PostgreSQLConnectionHandler.scala#L82-L104

[mauricio]

@sp-1234 none of the tests take this long to run: https://travis-ci.org/mauricio/postgresql-async/jobs/228756072 (other than the pool test, that has to timeout to it takes longer), not sure how this would be related to connection and be faster after the first since every connection will cause a new TCP connection to happen.

[sp-1234]

Maybe it's a platform specific issue, I were doing it on macOS. But this all is true story.

[njeuk]

@sp-1234 This is caused by an issue with osx DNS resolution, discussed here https://github.com/netty/netty/issues/6454 with resolutions https://stackoverflow.com/questions/39636792/jvm-takes-a-long-time-to-resolve-ip-address-for-localhost/39698914#39698914 . Adding my hostname explicitly to hosts worked well for me.