Hi, After have some more tests, I believe the things are like this:
1. client writes to ignite and timeout (client setting is 15ms) 2. client resets the connection since it seems dead (timeout). 3. server catches the connection reset and throw the exception Actually, it's just a normal use case about connection timeout. I guess "Too many open files" is caused by huge number and high frequency timeout and reconnecting. I tried to set nofile 491403 to avoid the problem. For 500 microseconds, it is just for testing. Generally, we use milliseconds level settings. Thanks. On Tue, Jul 3, 2018 at 9:57 PM, ilya.kasnacheev <ilya.kasnach...@gmail.com> wrote: > Hello! > > I have tried to reproduce your case, but I don't observe any growth of > number of open file descriptors on Ignite side. > > I think the problem here is on Go side. Please make sure to always close > connection if you open it. If your program is terminated, this is not so > strict, but if you create connections in a loop it can become a problem. > > Also, socket:[2322160] is not necessarily a UNIX socket, it is most often a > TCP socket as well. > > I also recommend changing '500 microseconds' to '500 milliseconds' because > there's not much you can expect to happen in half a millisecond or > one-2000th of a second. Especially when over a network. > > Regards, > > > > -- > Sent from: http://apache-ignite-users.70518.x6.nabble.com/
