tracker issue : CF-4060624

select a category, or use search below
(searches all categories and all time range)
Title:

new DataDirect 5.1.3 drivers leak Heap memory

| View in Tracker

Status/Resolution/Reason: Closed/Withdrawn/CannotReproduce

Reporter/Name(from Bugbase): DAN CORRIGAN / DAN CORRIGAN (DAN CORRIGAN)

Created: 09/21/2015

Components: Database

Versions: 10.0

Failure Type: Memory Leak

Found In Build/Fixed In Build: Final /

Priority/Frequency: Major / Some users will encounter

Locale/System: English / Win 2008 Server R2

Vote Count: 0

Problem Description: We had been running CF 10 Update 13 for over a year with no issues but after we upgraded to update 16 5 different busy servers of ours crashed every few days regularly.  With the help of Fusion-Reactor we could see Old Gen Heap space filling up with Full GCs collecting less and less until their was no more space in Old gen and we'd start getting Out of Memory errors and the CF would lock up.  We noticed new JDBC drivers were included with Update 14 and it was a big jump from DataDirect 4.1 to DataDirect 5.1.3

Steps to Reproduce: Have a busy server (ours have ~1.5 million page views a day) with multiple datasources (anywhere from 5 to 50) with 100+ websites in IIS and have maintain connections checked for each datasource

Actual Result:No memoery leak when using connection pooling

Expected Result:No memory leak when using connection pooling

Any Workarounds:Unchecking Maintain Connections for each data source seems to prevent the leak, but at a performance cost.

We had one server that did not crash after the update and the one thing we  found to be different on that server was that it had maintain connections unchecked on most datasources.  This was because this was an issue way back in CF7 or CF8 circa 2008 and those datasources had always been migrated over the years.  Unchecking maintain connections was the solution then too.  Newer JDBC drivers then corrected the issue.  I'm shocked to see that its back again after all this time.

----------------------------- Additional Watson Details -----------------------------

Watson Bug ID:	4060624

External Customer Info:
External Company:  
External Customer Name: DAN CORRIGAN
External Customer Email:  
External Test Config: CF10 x64 Update 16 

Server 2008 R2 x64

IIS 7.5

VMWare 5.5

8 VCPU 16 RAM

various heap sizes from 2GB to 6GB



We also had this same issue happen for a client we support that was running CF10 x32 Update 16 on Server 2003 with a s

Attachments:

Comments:

Forget to note I'm using SQL 2008 R2 databases
Comment by External U.
5809 | September 21, 2015 01:21:33 PM GMT
https://www.hass.de/content/coldfusion-10-macromedia-database-drivers-leaking-memory Another example of someone else having similar issue. I didn't have all his heap dump tools, but our graphs and results are about the same. This issue has now occured on another client running Update 16 and Server 2012 R2. The durarion until the OOM error depends on the amount of heap space and the busyness of the server, but ultimately they all have the same issue in the end.
Comment by External U.
5810 | September 23, 2015 10:35:09 AM GMT
So the two work around that work and prevent the servers from crashing within a few days are as follows: 1. uncheck maintain connections on all datasources. This works fine on servers where most websites have their own datasource. This does not work from a performance standpoint on servers that shared a datasource amongst hundreds of sites. For those I went with 2. Rollback the macromedia_drivers.jar (version 4.1.2) file to the version that was in Update 13 and prior. Simply replacing this file without any other changes restores server stability to what it was prior to Update 14.
Comment by External U.
5811 | October 05, 2015 07:36:25 AM GMT
To be clear, macromedia_drivers.jar 5.1.3 is the version that has the leak, and 4.1.2 is the version I rolled back to that fixed the issue.
Comment by External U.
5812 | October 05, 2015 07:39:40 AM GMT
Another update. The workaround of unchecking maintain connections helps but doesn't eliminate the problem. Instead of running out of OldGen space every few days, it makes it last a few weeks instead...but eventually it still does run out and crash CF. Rolling back to the older macromedia_driver.jar file has still been stable.
Comment by External U.
5813 | October 13, 2015 07:54:33 AM GMT
Dan, We have just released a ColdFusion 10 Update 20 with the latest database driver. Please check if this issue is resolved on your setup. Thanks, Nimit
Comment by Nimit S.
5814 | June 16, 2016 04:15:16 AM GMT
Dan, Have you tried this with the latest update? -Nimit
Comment by Nimit S.
5815 | June 23, 2016 12:21:06 AM GMT
Dan, Did you get a chance to look into this issue? -Nimit
Comment by Nimit S.
5816 | July 14, 2016 02:25:26 AM GMT
We are unable to repro this issue on the latest update released for ColdFusion 10. So, we are withdrawing this bug Please try it with the latest update and let us know? If you are still observing this issue. We will re-open this bug.
Comment by Nimit S.
5817 | July 27, 2016 11:11:34 PM GMT