8517 read error, hyperdex_client_reconfigure

[Gao Neal]

I'm trying to run a benchmark test on hyperdex warp. When I increase the number of threads to 35, reading failures comes a lot with hyperdex_client_reconfigure.
When I run 2 processes in two separate nodes, each with 30 threads, there's no hyperdex_client_reconfigure error. Any idea what might be going on?

Thanks

[Gao Neal]

it turns out that the clients are running out of open files. So increasing open file limits solved it.