I'm interested in what solution there is for this because I have the exact same problem but without parallel deployment.
[snip] On Mon, Nov 26, 2018 at 3:54 AM Gilles SCHLIENGER <gilles.schlien...@cncc.fr> wrote: > > Hi Christopher, > > Thanks for your email. > > About connection pools not being closed, maybe the connection pool is closed > but the connections to the database are not. > > Here are the tests I did: > > + TEST 1: > - I deploy my war > - I login to my webapp > - I check on my postgresql and mysql database that connections have been > opened (select...) > - I undeploy my webapp through the manager webapp > - I check that the connections are still opened > - Tomcat is still running > - I redeploy the exact samed webapp and login to my application > - No other connection is being opened > > + TEST 2: > - I deploy my war > - I login to my webapp > - I check on my postgresql and mysql database that connections have been > opened (select...) > - I deploy a new version of my application and login with another browser > - I check on my postgresql and mysql database that new connexions have been > opened (they doubled) > - I undeploy the first version of the webapp > - I check on my postgresql and mysql database and all connections are still > opened (even if I wait for a long time) > > For tests 1 and 2, I used C3p0, DBCP2 and even HikariCP > My results are the same. Using mysql, connection pools defined in server.xml (for user realm for access control to host manager app) and in my context.xml for my application. When I use the host-manager to reload an app, the connections are closed (no abandoned connection warnings) but not released. It is no until I stop tomcat completely and restart it that the connections are released in mysql. This has been the case for tomcat 7, 8.5 and 9 versions, with constant updates of mysql 5.7 and its driver, using both the apache connection pool and the tomcat connection pool. The driver lives in the tomcat/lib directory (since it is needed for the user realm datasource). > -----Message d'origine----- > De : Christopher Schultz [mailto:ch...@christopherschultz.net] > Envoyé : samedi 24 novembre 2018 17:19 > À : users@tomcat.apache.org > Objet : Re: Connection pool and parallel deployment problem > > -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA256 > > Gilles, > > On 11/23/18 05:07, Gilles SCHLIENGER wrote: [snip] > > > The warning/error messages are not actually linked to the > > connexions not closed. To ovoid these messages, you can: - move the > > jars (connexion pool, drivers...) into TOMCAT/lib - Have a > > ServletContextListener that calls > > AbandonedConnectionCleanupThread.checkedShutdown(); > > The Connector/J devs haven't been able to understand how ClassLoaders > work, and have never really fixed that long-standing bug in a > satisfying way as far as I know. But you should always use a > ServletContextListener to attempt to shut-down the > AbadonedConnectionCleanuopThread. > > BTW that thread does not work as documented. It's claimed that it > doesn't start unless you issue a query with a timeout, but it always > starts whether or not you use queries with timeouts. It's frustrating. > Should we be using checked or unchecked shutdown? --------------------------------------------------------------------- To unsubscribe, e-mail: users-unsubscr...@tomcat.apache.org For additional commands, e-mail: users-h...@tomcat.apache.org
