Bug #10
Cleaner Terminated abnormally when quering whole model
Start date:
Due date:
% Done:
0%
Estimated time:
Resolution:
fixed
Description
Mulgara crashes with the following output:
<br/>
<br/>
java.lang.Error: Cleaner terminated abnormally
<br/>
at sun.misc.Cleaner$1.run(Cleaner.java:130)
<br/>
at java.security.AccessController.doPrivileged(Native Method)
<br/>
at sun.misc.Cleaner.clean(Cleaner.java:127)
<br/>
at java.lang.ref.Reference$ReferenceHandler.run(Reference.java:124)
<br/>
Caused by: java.io.IOException: The process cannot access the file because anoth
<br/>
er process has locked a portion of the file
<br/>
at sun.nio.ch.FileChannelImpl.unmap0(Native Method)
<br/>
at sun.nio.ch.FileChannelImpl.access$000(FileChannelImpl.java:26)
<br/>
at sun.nio.ch.FileChannelImpl$Unmapper.run(FileChannelImpl.java:680)
<br/>
at sun.misc.Cleaner.clean(Cleaner.java:125)
<br/>
... 1 more
<br/>
<br/>
When the following process takes place:
<br/>
1. Start a new mulgara instance with an empty server1 directory
<br/>
2. Run a restore with our data (server 1 directory restores to ~6G)
<br/>
3. Run the following three querries one right after another
<br/>
-select $subject $predicate $object from <<a href="rmi://192.168.100.103/server1#data">rmi://192.168.100.103/server1#data</a>> where $subject $predicate $object;
<br/>
-select $subject $object from <<a href="rmi://192.168.100.103/server1#data">rmi://192.168.100.103/server1#data</a>> where $subject $predicate $object;
<br/>
-select $subject $predicate $object from <<a href="rmi://192.168.100.103/server1#data">rmi://192.168.100.103/server1#data</a>> where $subject $predicate $object;
<br/>
<br/>
The thrid query never returns. The command line where I started mulgara returns the java error described above.
<br/>
<br/>
Note: This does not happen on any of the pre-release versions of linux mulgara (Enterprise Redhat w/same specs).
Updated by Paula Gearon over 17 years ago
The only recent modification to Mulgara is the update to Lucene 2.0.0. This was done to help with a afile locking bug in an earlier version of Lucene. None of the Mulgara code explicitly locks files (to the best of my knowledge), so I believe this problem must be in Lucene somewhere.
<br/>
<br/>
Does the backed up data include any full text models?
Updated by Paula Gearon over 17 years ago
The only recent modification to Mulgara is the update to Lucene 2.0.0. This was done to help with a afile locking bug in an earlier version of Lucene. None of the Mulgara code explicitly locks files (to the best of my knowledge), so I believe this problem must be in Lucene somewhere.
<br/>
<br/>
Does the backed up data include any full text models?
<br/>
<br/>
A more likely problem is with Windows. A slightly older copy runs on Linux, while the newer one fails on Windows.
<br/>
- Does the older one work on Windows? (If so, then Lucene is probably the problem).
<br/>
- Does the newer version work on Linux? (If so, then Windows is probably the problem).
<br/>
<br/>
Windows file management is different in Windows, since the underlying file APIs have different guarantees on Windows to all other operating systems. This has caused problems in the past.
Updated by ben hysell - over 17 years ago
Sorry, no full text models in the code.
<br/>
<br/>
This issue has been present in every version of mulgara/kowari I have ever tested on an x64 machine. The same sequence of queries has brought them all down in the same manner with varrying degrees of results.
<br/>
<br/>
Older versions of kowari/mulgara would trash the server1 folder, (i.e. I would bring the system back up and no queries would complete). However with the latest 1.0 release the server1 directory seems intact and the system will operate. I brought the system down this morning after a restore, put it back up, and have been using it for the past two hours with no issues.
<br/>
<br/>
I have just recently obtained access to a linux machine for testing...the pre-release mulgara did not fail on the linux 64box. However I never had a chance to checkout kowari on a linux64 box.
<br/>
<br/>
Updated by ben hysell - over 17 years ago
Same happens when working from itql shell, only modification is adding a "limit 50" to the query so I'm not waiting all day for every triple to scroll by.
Updated by dmoll - over 17 years ago
I believe this issue is in Sun's bug tracker here:
<br/>
<br/>
<a href="http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=4938372">http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=4938372</a>
<br/>
<br/>
There is an issue with memory mapped files in Windows, haven't tested this on Linux or Mac OSX yet.
Updated by Andrae Muys - over 17 years ago
<a href="http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=4938372">http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=4938372</a> does appear to be the problem.
<br/>
<br/>
Reopen this bug if it is found on another platform.