Running on Ubuntu 16.04 LTS, database and clients run on the same system.

I am using the Dec2016 branch, and noticed that the database connections get dropped randomly, but in high frequency.

This error occurs irrespective for whether I use mclient, python package (pymonetdb) or the JDBC driver from the site.

From what I could figure if I use the forward=redirect option instead of the default proxy, the frequency reduces drastically, but is still present (if I leave the mclient connection idle for like 5 minutes it will drop - is there an idle time out somewhere ? Also is there a way to force a network connection locally instead of unix socket ? ).

Sample of some messages For the default proxy option

===========================

From merovingian.log

2016-12-01 10:59:33 MSG merovingian[29445]: target connection is on local UNIX domain socket, passing on filedescriptor instead of proxying

Python error due to dropped connection. 

JDBC Error (I am using PyCharm to execute some SQLs)

[08000] Connection to server lost! (mserver still alive?)

On the other hand, mclient does not show any error message, tried -X option, but no message when it dies. It just dies while executing a command (probably because the connection is dropped)

A peek at netstat shows that the connections disappear from it before mclient dies.

From merovingian.log

2016-12-01 11:22:59 MSG merovingian[29824]: redirecting client (local) for database 'voc' to mapi:monetdb://cerberus:50001/voc

Although for most part the connections are stable, if and when mclient dies, I do notice that the netstat entries are NOT being removed.

(see the entries piling up at the bottom though there are no active client processes)

Thanks,

Joseph