Hi everybody, I've created a data integration job to connect to MapR distribution I?ve two components, a tHDFSList and a TJava (to print the filename), you wan see it in the attached picture I configured the tHDFSList and ran the job but here is my problem : I think there is a memory leak? The java.exe process increase until 3Go and my computer crashes. I try without value in the different parameters, it?s the same behavior. I change the distribution and try with another : no problem, I have an error like ?server timeout?, but it?s normal. I don?t understand the problem with the MapR distribution. You can find the component screener in a attached picture. Please help ? For information, i had a first problem with Mapr, unsatisfiedlink error mapr client but i resolved it finding the solution in another topic (i had to instal a mapr client and declare the dll in the path environnement) Regards,
Hi, Thanks Ciaran for your reply. I'm using Win 7 and Java 1.7 I tried a Hive scenario (connection + input + close ) and it works fine. But when i try with tHDFS (or tPigLoad) i have the same problem, increase of the allocated memory and crash. Is it a problem with the MaprClient.dll ? I ask me that because I agree that it is a surprise that nobody reports the problem. Thank you for your help Regards,
Hi, I just try to change the MaprClient.dll. I have the memory leak with the MaprClient.dll form the mapr-client-2.0.0. I download he MaprClient.dll form the mapr-client-2.1.2 and i have a new error : Exception in thread "main" java.lang.Error: java.lang.Error: java.lang.UnsatisfiedLinkError: com.mapr.fs.MapRClient.initSpoofedUser(Ljava/lang/String;ILjava/lang/String;I)I
Is it a new problem with the version of the MaprClient.dll ?? Thank you
Hello, The second error is a mismatch between your mapr client and the maprfs library. Talend doesn't support MapR 2.1.2 yet. The first one comes from the dll as far as I know. Did you try to get help from MapR forums? Regards, Rémy.
Hello Rémy, Thanks for your reply. I understand for the problem with version 2.1.2. The memory leak with the MaprClient 2.0.0 is a MaprClient problem. I download the 1.2.0 MaprCLient (compatible with Talend) I don't have anymore the problem of the memory leak. But it's still not ok ... Now i have the error "Some error on socket 940 2013-03-20 14:36:55,3887 ERROR Cidcache fs/client/fileclient/cc/cidcache.cc:1047 Thread: -2 Lookup of volume mapr.cluster.root failed, error Unknown error(108), CLDB: *********:8020 trying another CLDB Some error on socket 940 2013-03-20 14:36:56,4768 ERROR Client fs/client/fileclient/cc/client.cc:226 Thread: -2 Failed to initialize client for cluster ****************:8020, error Unknown error(108) Exception in component tHDFSList_1 java.io.IOException: Could not create FileClient at com.mapr.fs.MapRFileSystem.lookupClient(MapRFileSystem.java:192) at com.mapr.fs.MapRFileSystem.lookupClient(MapRFileSystem.java:202) at com.mapr.fs.MapRFileSystem.listMapRStatus(MapRFileSystem.java:514) at com.mapr.fs.MapRFileSystem.listStatus(MapRFileSystem.java:558) at org.apache.hadoop.fs.FileSystem.listStatus(FileSystem.java:834) at org.apache.hadoop.fs.FileSystem.listStatus(FileSystem.java:859) at p_dtp_input.test_0_1.test.tHDFSList_1Process(test.java:443) at p_dtp_input.test_0_1.test.tHDFSConnection_1Process(test.java:362) at p_dtp_input.test_0_1.test.runJobInTOS(test.java:781) at p_dtp_input.test_0_1.test.main(test.java:649) "
Hello, This error means that your CLDB is not started because of a volume issue. Is your CLDB started on the cluster side? Are you able to browse this following web page: http://CLDB_HOSTNAME:8443 ? Rémy