[POOL-356] deadlock if borrowObject gets called to fast and maxIdle is 0

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 2.6.0
• Fix Version/s: 2.6.1
• Assignee: Mark Struberg
• Reporter: Mark Struberg
• Created: 24/Oct/18 10:50
• Updated: 25/Sep/19 20:14
• Resolved: 25/Sep/19 17:42

Description

I figured this while creating a unit test for OpenJPA. But also did see this in real production with commons-dbcp2. See DBCP-513 for more info.

See this comment for a precise explanation what happens https://issues.apache.org/jira/browse/DBCP-513?focusedCommentId=16660545&page=com.atlassian.jira.plugin.system.issuetabpanels%3Acomment-tabpanel#comment-16660545

The problem is basically that the logic to immediately destroy a pool object does not notify the DeLinkedQueue:

        if (isClosed() || maxIdleSave > -1 && maxIdleSave <= idleObjects.size()) {
            try {
                destroy(p);

But the borrowObject code is locking on that condition...

                    if (borrowMaxWaitMillis < 0) {
                        p = idleObjects.takeFirst();
                    } 

Issue Links

• blocks: DBCP-513 Hundreads of threads in Wait state with below stack trace
• breaks: POOL-376 invalidateObject should not return NullPointerException
• supercedes: POOL-327 GKOP: returnObject does not unblock threads waiting in borrowObject if maxIdle=0
• links to: GitHub Pull Request #11
• links to: GitHub Pull Request #25