Re: 2013 Errors and long Api Timeout (Aborted Cnx)
Hi, and thanx for your reply.
Our mysql network count around 1000 clients on LAN, and 2000 clients on WAN (ADSL 512/1024)
The problem occurs only on WAN, exactly as you understand it (Server abort the connexion, and client traps error 2013 around 7900 seconds later (7875=>7905 sec.)
Between ADSL msql clients and server, there is a firewall, but we are absolutely sure this is not the problem because :
- The Keep Alive timer is set to 3 hours
- No TCP frames are dropped during the non persistant connection from the client to the mysql server.
Does the problem can be described as below :
On ADSL Network, the connection lives during 24 hours, and is re-initialized exactly during a query (Select, insert, ....)
The Server detects correctly the breaking connection, trace the error, and send a notification to client that the connection is broken.
Now, can we imagine that the client does not detect the breaking connection, cant receive the notification from the server (because the connection si broken), is pending 2 hours until it reconnects, retries or something in the application that reinit the connection.
How does the mysql APIs works when this problems happens ?
Is there a such application timeout inside the APIs (mysql_real_query(), ...) that may explain this too long timer ?
Thanx to help me, because is a really big problem, and im pretty sure this is an applicative MySql API problem.
-----------------------------------
B. Lepoittevin
MySql DBA - Network Administrator