Just another data point on this. Another reason I don't think Hikari is getting initialized. Note the pool name " HikariPool-9 - is starting". I always give my pools an explicit name, and the log message is using a Hikari default name.
2018-02-23T09:28:12,725 | INFO | CM Configuration Updater (Update: pid=org.ops4j.datasource.787a9126-3325-47e5-a5a9-8c7ad43f0ed8) | MultiServiceTracker | 56 - org.ops4j.pax.jdbc.config - 1.1.0 | Found service org.ops4j.pax.jdbc.pool.common.PooledDataSourceFactory with filter (&(objectClass=org.ops4j.pax.jdbc.pool.common.PooledDataSourceFactory)(pool=hikari)(xa=false)) 2018-02-23T09:28:12,725 | INFO | CM Configuration Updater (Update: pid=org.ops4j.datasource.787a9126-3325-47e5-a5a9-8c7ad43f0ed8) | DataSourceRegistration | 56 - org.ops4j.pax.jdbc.config - 1.1.0 | Found DataSourceFactory. Creating DataSource bam/ds/sap/sidecar 2018-02-23T09:28:12,778 | INFO | CM Configuration Updater (Update: pid=org.ops4j.datasource.787a9126-3325-47e5-a5a9-8c7ad43f0ed8) | HikariDataSource | 145 - com.zaxxer.HikariCP - 2.4.1 | HikariPool-9 - is starting. Regards, Scott -----Original Message----- From: Jean-Baptiste Onofré [mailto:j...@nanthrax.net] Sent: Monday, February 19, 2018 11:21 PM To: email@example.com Subject: Re: Pax connection pools Hi Scott, let me take a look. It could be related to pax-jdbc-config and the related features declaration. Regards JB On 20/02/2018 01:36, Leschke, Scott wrote: > I've emailed about this before but some time ago I mentioned that I > didn't think that the PAX connection pools were initializing the > underlying CP implementation (Hikari in my case) properly. In > particular, connections are getting dropped after 30 mins (the Hikari > default) even though connection maxLifetime is set to inifinite (0) > and idleTimeout is set to 8 hrs. > > Christian sent me a link to the PAX Hikari initialization code which > certainly looked correct. This lead me to believe that the problem was > not with PAX but the Hikari version that is bundled with it. While I'm > not positive this is the case, I suspect it is. > > Might it be possible for PAX to log the pool implementations > configuration AFTER it creates the pool? Also, will the version of > Hikari be updated in the next release of Karaf? The latest version > appears to be 2.7.7. I believe the version that comes with the most > recent version of Karaf is still at 2.4.1. > > Regards, > > Scott >