Improper handling of lost database connection
---------------------------------------------
Key: GEOS-1107
URL: http://jira.codehaus.org/browse/GEOS-1107
Project: GeoServer
Issue Type: Bug
Components: PostGIS
Affects Versions: 1.4.1
Environment: GeoServer 1.4.1
O.S.: Windows Server 2003 with SP1
Web Server: IIS
Servlet Container: Jetty
PostgreSQL/PostGIS 8.0.x database backend on Linux RHE 4.0
Reporter: Steve Marshall
Assignee: Andrea Aime
Fix For: 1.5.2
GeoServer does not properly handle a lost connection to a PostgreSQL/PostGIS database. GeoServer does not detect if the connection goes bad until after it has executed a failed query. The error handling seems to get GeoServer back in the proper state, but at a cost: the query that caused the error handling logic to execute returns as a failure.
To test this error, you can do any thing that will cause a database connection to go bad, then check the error return from a WFS query that uses this database connection. The error message will look something like this:
500 Server closed connection without sending any data back
Suggestions for causing the connection to go bad include:
1. Keep the connection idle until the TCP keepalive feature closes the connection. For default system and PostgreSQL configurations, this will take about 2 hours 12 minutes.
2. kill the postgres backend process associated with the GeoServer application. Use the PostgreSQL system view pg_stat_activity to determine which backend is associated with the GeoServer application based on DB user id and the client host IP address.
More details on a case that revealed this error can be found in this GeoServer user group link:
http://www.nabble.com/null-value-in-log-file-tf3680605.html#a10789565
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators: http://jira.codehaus.org/secure/Administrators.jspa
-
For more information on JIRA, see: http://www.atlassian.com/software/jira