Oracle Connector on Solaris - memory leak?
Moderators: chulett, rschirm, roy
Oracle Connector on Solaris - memory leak?
Hi Experts,
During one of my recent DataStage assignments I observed strange behavior of Oracle Connector stage - it consumes as much of heap memory as the size of accessed data is.
I use Oracle Connector for reading data from one of Oracle 10.2.0.4 tables, all the array/transaction size properties are default (2000 records). Oracle Client version is 10.2.0.4 too, both server/client sides are Sun Solaris 10.
Is this possible that DataStage intentionally buffers all fetched rows before it finishes reading data?
It's worth to add the next operation after Oracle Connector is Filter that distributes data to two Sort stages, however according to job score each of this stages has it's own operator assigned.
Regards
Buzz
During one of my recent DataStage assignments I observed strange behavior of Oracle Connector stage - it consumes as much of heap memory as the size of accessed data is.
I use Oracle Connector for reading data from one of Oracle 10.2.0.4 tables, all the array/transaction size properties are default (2000 records). Oracle Client version is 10.2.0.4 too, both server/client sides are Sun Solaris 10.
Is this possible that DataStage intentionally buffers all fetched rows before it finishes reading data?
It's worth to add the next operation after Oracle Connector is Filter that distributes data to two Sort stages, however according to job score each of this stages has it's own operator assigned.
Regards
Buzz
I'm not sure anyone here could actually answer that unless they've ask this same question of IBM. Unless someone else pipes up, I'd suggest you ask your official support provider about that, see if there are any known issues with the connector in your flavor/version. Then let us know what you find out.
BTW, what 8.x version are you running?
BTW, what 8.x version are you running?
-craig
"You can never have too many knives" -- Logan Nine Fingers
"You can never have too many knives" -- Logan Nine Fingers
I managed to replicate this issue on my internal AIX server (note it's not Solaris but issue is there too!), unfortunately installing newer version of Oracle Client didn't help - no matter if it's Oracle Client 10.2.0.1 or Oracle Client 11.2.0.1 I still observe memory leak.
Is there a chance Oracle fixed this issue in 10.2.0.5 version and then introduced this 'bug' again? Somehow I don't believe it![Confused :?](./images/smilies/icon_confused.gif)
Is there a chance Oracle fixed this issue in 10.2.0.5 version and then introduced this 'bug' again? Somehow I don't believe it
![Confused :?](./images/smilies/icon_confused.gif)
Actually I spent 10 years with Oracle, but mostly playing with in-database processing, with no serious usage of OCI from outside ![Wink :wink:](./images/smilies/icon_wink.gif)
I agree Oracle is full of bugs (as each software BTW) but this issue sounds like a really straightforward one, that should emerge really quickly in any QA test...
Anyway, thanks for your comments![Smile :)](./images/smilies/icon_smile.gif)
![Wink :wink:](./images/smilies/icon_wink.gif)
I agree Oracle is full of bugs (as each software BTW) but this issue sounds like a really straightforward one, that should emerge really quickly in any QA test...
Anyway, thanks for your comments
![Smile :)](./images/smilies/icon_smile.gif)
I was pulling your leg, of course.
So... you've marked this as Resolved but we really didn't come to any true resolution from what I can tell. Can you let us know if you have actually resolved this, please, and what that resolution was?
![Wink :wink:](./images/smilies/icon_wink.gif)
So... you've marked this as Resolved but we really didn't come to any true resolution from what I can tell. Can you let us know if you have actually resolved this, please, and what that resolution was?
-craig
"You can never have too many knives" -- Logan Nine Fingers
"You can never have too many knives" -- Logan Nine Fingers
For all interested in fixing this issue on AIX platform - today IBM support officially published an APAR:
http://www-01.ibm.com/support/entdocvie ... wg1JR37402
Cheers
Buzz
http://www-01.ibm.com/support/entdocvie ... wg1JR37402
Cheers
Buzz