HashedWheelTimer exception when running tests in parallel

[Christian Kaps]

Hi,

I have an app which runs a lot of test against an embedded MongoDB in parallel. Now I get an HashedWheelTimer exception and some tests failed.

p.s.a.i.n.u.HashedWheelTimer - You are creating too many HashedWheelTimer instances. HashedWheelTimer is a shared resource that must be reused across the JVM,so that only a few instances are created.

ReactiveMongo uses the HashedWheelTimer: https://github.com/ReactiveMongo/ReactiveMongo/blob/f63f429b2eedc18fab565565b1c58e9a60a30703/driver/src/main/scala/core/nodeset/ChannelFactory.scala#L58

Is this an bug?

Best regards,

Christian

[Cédric Chantepie]

The ChannelFactory is managed by MongoConnection, which is a pool. According that design it's not expected to create so many pool that it raise the issue with Netty, as the pool can itself be used for several parallel tests.

[Christian Kaps]

Any idea how I could solve that with Play and DI? Play creates a new application for every test, so I have no idea how I could provide a shared pool between the tests.

[Cédric Chantepie]

Not using such DI or executing more tests per injected app.