I am dealing with what I think is a connection pooling issue. The
initial trigger was that we have an external service that calls a
Liferay API (7.3.2-ga3 CE) . The external service indicates that
Liferay failed to respond from time to time. Now, I understand that
there are many moving parts here but my primary focus is to make sure
that the LR instance is optimally configured. So, my initial
assumption was that we had a DB connection pool issue. We have a
connection to an external SQL server that originally was using the
default Tomcat DB connection pool. Using David Nebinger's blog post,
I switched over to using Hikari. That doesn't seem to have helped much.
To test the pooling and responses, I added a JMeter test to
crank up 200 requests to the API. I also set the Hikari logging to
DEBUG. What I see in the log is that HikariPool-1 throws an error:
2021-01-18 15:03:28.499 ERROR
[http-nio-8080-exec-193][JSONWebServiceServiceAction:114]
org.hibernate.exception.GenericJDBCException: Cannot open
connection
2021-01-18 15:03:28.499 ERROR
[http-nio-8080-exec-192][JDBCExceptionReporter:234] HikariPool-1 -
Connection is not available, request timed out after 30000ms.
I think that indcates that the pool is exhausted (not sure
though). So my initial question here is what is HikariPool-1 ? I
see HikariPool-1, HikariPool-2 and HikariPool-3 listed in the DEBUG
log. I assume that one of them is the MSSQL server connection,
probably HikariPool-3. I understand that by default, Liferay uses
Hikari, so I expect that either HikariPool-1 or HikariPool-2 is the
Liferay connection. But why are there two?
Further, if HikariPool-1 and HikariPool-2 are related to
Liferay, where do I configure those connection properties since at
least one of them seems too small? Any pointers here on how best to
configure the Hikari pools would be helpful. That, and helping me to
determine why there seem to be two pools for Liferay OOTB.
Thanks
