I'm testing out the parallel deployment process for my company and am seeing a difference in how user migrations to the new version of the application occur based upon how you deploy your application. Our configuration involves many different web sites that all point to the same folder. Some example context configuration files are provided below to help illustrate what I mean:
Website1.xml <Context reloadable="true" crossContext="true" sessionCookiePath="/" docBase="/Users/jeremy/tomcat/manualDeploy/myapp"></Context> Website2.xml <Context reloadable="true" crossContext="true" sessionCookiePath="/" docBase="/Users/jeremy/tomcat/manualDeploy/myapp"></Context> Website3.xml <Context reloadable="true" crossContext="true" sessionCookiePath="/" docBase="/Users/jeremy/tomcat/manualDeploy/myapp"></Context> Yesterday some individuals from this mailing list helped me understand that if I create a new context file with ##001 appended at the end then that will allow me to deploy a new version of the application. For example: Website1##002.xml <Context reloadable="true" crossContext="true" sessionCookiePath="/" docBase="/Users/jeremy/tomcat/manualDeploy/myapp-v2"></Context> When I run Jmeter requests consistently against website1 and then deploy the new version of the web site (website1##002) I do not see any active sessions for website##002 in tomcat manager even after website1##002 is fully loaded. Please note that if I open up a new safari browser window and access website1#002 then I do see an active session. What is confusing me at this point is that if I deploy a WAR using the auto deployment method (copying the myapp.war into the webapps folder) and then deploy a new version using the same mechanism (myapp##002.war) then I do see the users automatically migrated to the new version when running my Jmeter tests. Please note that this behavior (immediate migration) is what we desire. In addition to the fact that when the migration of users occurs is different, I am also seeing a huge discrepancy between the number of sessions that tomcat manager says are currently active when I deploy using the folder approach vs. the automatic deployment approach. When using the folder approach the number of sessions always matches the number of users (via a thread group) in Jmeter. However, when I use the automatic deployment process the number of sessions continues to grow as I allow Jmeter to run longer (at one point I had thousands of sessions). Can anyone explain why I'm seeing a difference in the behavior between the two deployment mechanisms and if so, how would I mimic what is happening in regards to the automatic migration of users to new versions while using the folder deployment approach. Thank you very much for your assistance, Jeremy
