Update notes for 2017-12-20

content/post/2017-12.md

@@ -423,18 +423,18 @@ $ schedtool -D -e ionice -c2 -n7 nice -n19 dspace index-discovery
 - Then I create a global `Resource` in the main Tomcat _server.xml_ (inside `GlobalNamingResources`):
 
 ```
-	<Resource name="jdbc/dspace" auth="Container" type="javax.sql.DataSource"
-		  driverClassName="org.postgresql.Driver"
-		  url="jdbc:postgresql://localhost:5432/dspace"
-		  username="dspace"
-		  password="dspace"
-          initialSize='5'
-          maxActive='50'
-          maxIdle='15'
-          minIdle='5'
-          maxWait='5000'
-          validationQuery='SELECT 1'
-          testOnBorrow='true' />
+<Resource name="jdbc/dspace" auth="Container" type="javax.sql.DataSource"
+	  driverClassName="org.postgresql.Driver"
+	  url="jdbc:postgresql://localhost:5432/dspace"
+	  username="dspace"
+	  password="dspace"
+      initialSize='5'
+      maxActive='50'
+      maxIdle='15'
+      minIdle='5'
+      maxWait='5000'
+      validationQuery='SELECT 1'
+      testOnBorrow='true' />
 ```
 
 - Most of the parameters are from comments by Mark Wood about his JNDI setup: https://jira.duraspace.org/browse/DS-3564
@@ -623,3 +623,30 @@ javax.naming.NoInitialContextException: Need to specify class name in environmen
 - Oh cool! `select * from pg_stat_activity` shows "PostgreSQL JDBC Driver" for the application name! That's how you know it's working!
 - If you monitor the `pg_stat_activity` while you run `dspace database info` you can see that it doesn't use the JNDI and creates ~9 extra PostgreSQL connections!
 - And in the middle of all of this Linode sends an alert that CGSpace has high CPU usage from 2 to 4 PM
+
+## 2017-12-20
+
+- The database connection pooling is definitely better!
+
+![PostgreSQL connection pooling on DSpace Test](/cgspace-notes/2017/12/postgres-connections-week-dspacetest.png)
+
+- Now there is only one set of idle connections shared among all the web applications, instead of 10+ per application
+- There are short bursts of connections up to 10, but it generally stays around 5
+- Test and import 13 records to CGSpace for Abenet:
+
+```
+$ export JAVA_OPTS="-Dfile.encoding=UTF-8 -Xmx512m -XX:+TieredCompilation -XX:TieredStopAtLevel=1"
+$ dspace import -a -e aorth@mjanja.ch -s /home/aorth/cg_system_20Dec/SimpleArchiveFormat -m systemoffice.map &> systemoffice.log
+```
+
+- The fucking database went from 47 to 72 to 121 connections while I was importing so it stalled.
+- Since I had to restart Tomcat anyways, I decided to just deploy the new JNDI connection pooling stuff on CGSpace
+- There was an initial connection storm of 50 PostgreSQL connections, but then it settled down to 7
+- After that CGSpace came up fine and I was able to import the 13 items just fine:
+
+```
+$ dspace import -a -e aorth@mjanja.ch -s /home/aorth/cg_system_20Dec/SimpleArchiveFormat -m systemoffice.map &> systemoffice.log
+$ schedtool -D -e ionice -c2 -n7 nice -n19 dspace filter-media -i 10568/89287
+```
+
+- The final code for the JNDI work in the Ansible infrastructure scripts is here: https://github.com/ilri/rmg-ansible-public/commit/1959d9cb7a0e7a7318c77f769253e5e029bdfa3b