Details
-
Bug
-
Resolution: Fixed
-
High
-
6.0.0-rc5
-
Severity 2 - Major
-
Description
After about an hour or two I tried to edit another page in our Confluence 6.0 rc5 test instance.
I got an error when trying to edit the page. This seems to be a timeout issue with our JDBC JNDI pool used for communicating with the Confluence database. I attach a combined log of atlassian-confluence.log, atlassian-synchrony.log and MySQL server log.
synchrony-communication-error.log
The MySQL server seems to abort a connection after some time because there was no response to data delivered to the synchrony client. I made sure that the MySQL server was running. There was definitely no connection problem to the server as several page where retrieved and open in Confluence.
Is this due to a misconfiguration or a general error in Synchrony?
When I retried editing a page another hour later, the problem didn't exist any more. Might be an issue with reusing a database connection which has been idle/closed when not being used for some time.
Regards,
Jan-Peter
Solution:
We will make the connection pool's parameters of Synchrony configurable via the system properties. Basically, a Confluence administrator can set the following properties:
- synchrony.database.test.connection.on.checkin (default: true)
- synchrony.database.idle.connection.test.period (defualt: 1800 seconds, or 30 minutes)
- synchrony.database.max.idle.time.excess.connections (default: 1800 seconds, or 30 minutes)
- synchrony.database.max.idle.time (default: 10800 seconds, or 3 hours)