Tuesday, October 12, 2004
Weblogic and the curse of the Session Id
While implementing a simple session time out scenario , i encountered problems which seemed bizarre .. i discovered that multiple sessions being created .Strangely i saw that these sessions were being created even though an existing session was in place .and clicking any link on the site created a new session almost everytime . Deciding to do something different i entered the url on the browser and voila !! no sessions were being created ... Further investigation led me to the weblogic.xml which had an encode sessionidinquery params value set to true . What this triggered was the weblogic now only understood if there was jsessionId present in the browser .and the web site did'nt use any of the response.encodeUrl() so clicking on any link on the website resulted in creating a new session.
Changing the parameter encodeSessionIdinQueryParams to false restored some sanity and able to carryout the timeout implementation.
While implementing a simple session time out scenario , i encountered problems which seemed bizarre .. i discovered that multiple sessions being created .Strangely i saw that these sessions were being created even though an existing session was in place .and clicking any link on the site created a new session almost everytime . Deciding to do something different i entered the url on the browser and voila !! no sessions were being created ... Further investigation led me to the weblogic.xml which had an encode sessionidinquery params value set to true . What this triggered was the weblogic now only understood if there was jsessionId present in the browser .and the web site did'nt use any of the response.encodeUrl() so clicking on any link on the website resulted in creating a new session.
Changing the parameter encodeSessionIdinQueryParams to false restored some sanity and able to carryout the timeout implementation.
Monday, October 11, 2004
The Phantom data trail
For those who have come in late ..
We were happy to use the connection pool for database access . just get the connection from the pool use it and release it back... life's cool until one day when we decided to execute a transaction . we followed the following steps ...
For those who have come in late ..
We were happy to use the connection pool for database access . just get the connection from the pool use it and release it back... life's cool until one day when we decided to execute a transaction . we followed the following steps ...
- get a connection from the free pool
- set the connection's auto commit property to false (needed for a transaction
- put the db execute/update statements within a try catch block
- In the catch block log the exception and,close the statement release the connection and throw the exception back
- end of the method con.commit()
- release the connection back to the pool
This worked pretty much fine for all the happy flows ... whenever an error happened we used to have some data appearing in the database long time after the operation was over .The mystery further deepened when we found out this data which was appearing was due to failed transaction.
Reality dawned a li'l later when I found out that in the catch block one crucial step was missing conn.rollback() w/o which we used to release it to the pool and when someone else gets this same connection and performs certain other operations along with those operations whatever statements were successful in the previous operations were being updated
Moral of the story
pool or no pool ... to keep life a li'l simpler and easier follow the best practices
Subscribe to:
Posts (Atom)