Description:
When experiencing connectivity problems between a client and a mysql server it can often be very hard to diagnose the EXACT reason for these problems. There are 2 status variables which provide information (see http://dev.mysql.com/doc/refman/5.6/en/server-status-variables.html) and these are:
Aborted_clients
Aborted_connections
yet reading the associated text it becomes obvious that the number of different reasons for these errors is quite varied:
Aborted_clients (3 different reasons stated)
Aborted_connections (4 specific reasons stated)
There are also some common reasons for issues.
So if you need to diagnose an occasional problem reaching a server the statistics there are usually not as complete as one would like. The problem may not be the mysql server at all but a network, firewall, client (software), or client (behaviour) issue, but this is not clear.
There are hundreds of general questions and queries about things like "MySQL has gone away" when a client loses a connection but there is also a better need to determine if a failed connection (especially one with a connect timeout) fails as diagnosing this type of error, especially if it is very intermittent is really hard, and traffic sniffing hours of "success connections" to look for the single failure is really hard.
So I think more detailed statistics of "unclean" connects or disconnects would be most useful.
How to repeat:
Make lots of connects to a mysql server and intermittently trigger problems by doing things like disconnect the network, or kill the connecting clients. Now try to figure out some of the different reasons for these unclean connects based on the status variables. There is some logging but there's not much detail in the way of the status variables. Also logging provides an attack vector which could fill up a log file so it's often not good to log too much (under normal circumstances).
Suggested fix:
It would be most useful if more status variables could be added which would include the appropriately more detailed reasons (maintaining the current ones for backwards compatibility) so that diagnosis of issues like this could be made much better.
The actual reporting of errors to the client could also be improved to help identify the causes of certain issues. This may not necessarily require the error codes to change but simply the error text providing a more specific description of the cause.
