Possible deadlock during populate internal about bogus

[bchavez]

Hey @Mpdreamz ,

I think you're right. I was able to reproduce the issue with the elasticsearch repo. Seems like there was quite a few static field Faker initializations that were being fired up at the same time.

I changed the locker to ExecutionAndPublication and it seemed to solve the issue after running the elasticsearch unit tests a few times now.

Please feel free to re-open the issue if you run into the issue again.

Bogus v8.0.2 with the ExecutionAndPublication should available soon as it gets indexed shortly.

Thanks,
Brian

[bchavez]

Hi @Mpdreamz ,

Sorry about the trouble you're having. I wrote a few unit tests to try to reproduce a deadlock; however, I'm not able to reproduce the behavior you're seeing.

Is this in the elasticsearch-net repo? Is it easy to run unit tests? Is there anything special I need to run the unit tests?

Thanks,
Brian

[Mpdreamz]

Hi @bchavez

It is yeah, you should be able to clone it and checkout the https://github.com/elastic/elasticsearch-net/tree/5.0.0-rc4 tag and run build from the commandline. This will build and run the unit tests.

The deadlock does not happen always, as deadlocks tend to do. It doesn't even happen often enough for us to get into gear but when it does it tends to happen constantly during that period.

I do think having the lazy execute with ExecutionAndPublication be a safer option so no synchronization has to happen on the winning threads returned object that everyone wants to lock on.

[Mpdreamz]

Awesome thanks Brian!