Hi all,
I am trying to trace down a problem I have with logging.
The problem is that on the BI Server I am logging out to the PENTAHOFILE and CONSOLE appenders (I have not changed the config from the sourceforge download). I find after a certain amount of time the pentaho.log file no longer contains entries but catalina.log continues happily and contains warnings like
log4j:ERROR Attempted to append to closed appender named [PENTAHOFILE].
Restarting the tomcat fixes it, but it re-occurs regularly. After some trial and error I believe it is being caused by running any report which uses kettle as a data source (since I had reports of this nature scheduled through kettle transforms the logging would break regularly).
I can get around this, by using scheduled kettle jobs to import data to a common location and then change the report to use SQL over that location rather than the kettle source over multiple data sources but it would be nice to understand the problem.
Reading around the subject I understand that kettle has its own log4j file in the jar and logging classes which build on log4j (new levels etc) and override some of the behavior defined in the log4j file including forcing a console appender if one cannot be found.
I don't know how to go about proving whether this is right or wrong but my best guess is that for some reason the LogWriter is deciding an appropriate appender cannot be found, creates a new console appender (hence the output to Catalina.out continues happily) and the PENTAHOFILE appender is lost somewhere.
Can anyone with more experience of this suggest anything?
Thanks,
James
I am trying to trace down a problem I have with logging.
The problem is that on the BI Server I am logging out to the PENTAHOFILE and CONSOLE appenders (I have not changed the config from the sourceforge download). I find after a certain amount of time the pentaho.log file no longer contains entries but catalina.log continues happily and contains warnings like
log4j:ERROR Attempted to append to closed appender named [PENTAHOFILE].
Restarting the tomcat fixes it, but it re-occurs regularly. After some trial and error I believe it is being caused by running any report which uses kettle as a data source (since I had reports of this nature scheduled through kettle transforms the logging would break regularly).
I can get around this, by using scheduled kettle jobs to import data to a common location and then change the report to use SQL over that location rather than the kettle source over multiple data sources but it would be nice to understand the problem.
Reading around the subject I understand that kettle has its own log4j file in the jar and logging classes which build on log4j (new levels etc) and override some of the behavior defined in the log4j file including forcing a console appender if one cannot be found.
I don't know how to go about proving whether this is right or wrong but my best guess is that for some reason the LogWriter is deciding an appropriate appender cannot be found, creates a new console appender (hence the output to Catalina.out continues happily) and the PENTAHOFILE appender is lost somewhere.
Can anyone with more experience of this suggest anything?
Thanks,
James