Topic
  • 11 replies
  • Latest Post - ‏2014-08-11T17:29:51Z by 7PJN_Martin_Mašek
7PJN_Martin_Mašek
7PJN_Martin_Mašek
17 Posts

Pinned topic Installation BI EE 3.0 - Failed to execute hive commands GRANT

‏2014-08-05T19:41:19Z |

Hi community,

I'm installing BI 3.0 EE with adaptive MapReduce and GPFS. Linux version is RHEL 6.5. I've set everything up and checker utility passed. Installation also passed but after check (or whatever it is, the second bar after installation) finished errors were shown. I examined the log and found that first error was Failed to execute hive commands GRANT on bivm (bivm our machine). I have found no solution on the google and there is no troubleshooting in documentation or some more info in the error message :/ I would appreciate some help.

Martin

  • Tuong_Truong
    Tuong_Truong
    2 Posts
    ACCEPTED ANSWER

    Re: Installation BI EE 3.0 - Failed to execute hive commands GRANT

    ‏2014-08-06T02:36:10Z  

    Hi, I found that post you are talking about but there was no answer.

    Here are the files.

    Hi Martin,

    The Hive grant error is the result of the catalog service is not properly installed.  In the install log, this error message indicate a missing library that DB2 needs., You need to install that on the chosen catalog node, and reinstall:

     

    [ERROR] DeployManager - catalog failed
    com.ibm.xap.mgmt.DeployException: DB2 start service failed -- bivm=1 (exit code: 127 -- ssh -o StrictHostKeyChecking=no bivm sudo -u catalog -i db2start
    db2start: error while loading shared libraries: libnuma.so.1: cannot open shared object file: No such file or directory
    )
    at com.ibm.xap.mgmt.database.DB2Operator.startInstance(DB2Operator.java:351)
    at com.ibm.xap.mgmt.commoncatalog.DB2CatalogDeployer.startService(DB2CatalogDeployer.java:374)
     

     

    You can find more info on this library requirement in the BI 3.0 release note: 

    http://www-01.ibm.com/support/knowledgecenter/SSPT3X_3.0.0/com.ibm.swg.im.infosphere.biginsights.product.doc/doc/release_notes.html

    Tuong

    Updated on 2014-08-06T16:37:55Z at 2014-08-06T16:37:55Z by Tuong_Truong
  • tfriedr
    tfriedr
    29 Posts

    Re: Installation BI EE 3.0 - Failed to execute hive commands GRANT

    ‏2014-08-05T19:58:03Z  

    Hi Martin,

    We need a few more details to analyze the problem. Is it the same error message as described in this question and do you see any other errors in the log?
    https://www.ibm.com/developerworks/community/forums/html/topic?id=46a61b33-3917-48a4-87fb-e3e25c88e958&ps=25
    Can you provide the install log and the fullinstall.xml?

    Thomas

  • 7PJN_Martin_Mašek
    7PJN_Martin_Mašek
    17 Posts

    Re: Installation BI EE 3.0 - Failed to execute hive commands GRANT

    ‏2014-08-05T20:23:29Z  

    Hi, I found that post you are talking about but there was no answer.

    Here are the files.

    Attachments

  • 7PJN_Martin_Mašek
    7PJN_Martin_Mašek
    17 Posts

    Re: Installation BI EE 3.0 - Failed to execute hive commands GRANT

    ‏2014-08-05T20:53:02Z  

    And my second problem is that now, after the "installation", I can't change things and install again because of error

    ERROR] (CDYIN0323E) bivm.com: The ~catalog/sqllib directory already exists. This will cause the DB2 installer to fail. Remove this directory and run the utility again.

    and these kinds of errors (another port numbers)

    ERROR] (CDYIN0381E) bivm.com: Port 62000 conflicts with an existing service port. Change to another port and run this utility again.

    I rm -rf ~catalog/sqllib but it didn't help and netstat |grep 62000 returned nothing.

    Updated on 2014-08-05T20:56:57Z at 2014-08-05T20:56:57Z by 7PJN_Martin_Mašek
  • Tuong_Truong
    Tuong_Truong
    2 Posts

    Re: Installation BI EE 3.0 - Failed to execute hive commands GRANT

    ‏2014-08-06T02:36:10Z  

    Hi, I found that post you are talking about but there was no answer.

    Here are the files.

    Hi Martin,

    The Hive grant error is the result of the catalog service is not properly installed.  In the install log, this error message indicate a missing library that DB2 needs., You need to install that on the chosen catalog node, and reinstall:

     

    [ERROR] DeployManager - catalog failed
    com.ibm.xap.mgmt.DeployException: DB2 start service failed -- bivm=1 (exit code: 127 -- ssh -o StrictHostKeyChecking=no bivm sudo -u catalog -i db2start
    db2start: error while loading shared libraries: libnuma.so.1: cannot open shared object file: No such file or directory
    )
    at com.ibm.xap.mgmt.database.DB2Operator.startInstance(DB2Operator.java:351)
    at com.ibm.xap.mgmt.commoncatalog.DB2CatalogDeployer.startService(DB2CatalogDeployer.java:374)
     

     

    You can find more info on this library requirement in the BI 3.0 release note: 

    http://www-01.ibm.com/support/knowledgecenter/SSPT3X_3.0.0/com.ibm.swg.im.infosphere.biginsights.product.doc/doc/release_notes.html

    Tuong

    Updated on 2014-08-06T16:37:55Z at 2014-08-06T16:37:55Z by Tuong_Truong
  • Tuong_Truong
    Tuong_Truong
    2 Posts

    Re: Installation BI EE 3.0 - Failed to execute hive commands GRANT

    ‏2014-08-06T16:49:42Z  

    And my second problem is that now, after the "installation", I can't change things and install again because of error

    ERROR] (CDYIN0323E) bivm.com: The ~catalog/sqllib directory already exists. This will cause the DB2 installer to fail. Remove this directory and run the utility again.

    and these kinds of errors (another port numbers)

    ERROR] (CDYIN0381E) bivm.com: Port 62000 conflicts with an existing service port. Change to another port and run this utility again.

    I rm -rf ~catalog/sqllib but it didn't help and netstat |grep 62000 returned nothing.

    Martin,

    Did you run a proper uninstall - uninstall.sh using your biadmin login?   These issue should be taken care of by the uninstaller.

    In order to remove of DB2 residual artifacts manually, you need to remove a couple of things:

    1. Remove the sqllib directory - for catalog user (on catalog node) as well as BigSQL userid on all other nodes

    2. On each node DB2 was installed (practically all node), remove all files under  /var/db2/

    3. Edit /etc/services file and remove any DB2 service port for catalog and BigSQL on all nodes.

        should look something like:

    db2c_catalog    xxxxx/tcp
    DB2_bigsql      xxxxx/tcp
    DB2_bigsql_END  xxxxxtcp
    db2c_bigsql     xxxxxtcp

     

    Hope this will get you moving again..

    Tuong
     

  • 7PJN_Martin_Mašek
    7PJN_Martin_Mašek
    17 Posts

    Re: Installation BI EE 3.0 - Failed to execute hive commands GRANT

    ‏2014-08-06T19:02:58Z  

    Hi Martin,

    The Hive grant error is the result of the catalog service is not properly installed.  In the install log, this error message indicate a missing library that DB2 needs., You need to install that on the chosen catalog node, and reinstall:

     

    [ERROR] DeployManager - catalog failed
    com.ibm.xap.mgmt.DeployException: DB2 start service failed -- bivm=1 (exit code: 127 -- ssh -o StrictHostKeyChecking=no bivm sudo -u catalog -i db2start
    db2start: error while loading shared libraries: libnuma.so.1: cannot open shared object file: No such file or directory
    )
    at com.ibm.xap.mgmt.database.DB2Operator.startInstance(DB2Operator.java:351)
    at com.ibm.xap.mgmt.commoncatalog.DB2CatalogDeployer.startService(DB2CatalogDeployer.java:374)
     

     

    You can find more info on this library requirement in the BI 3.0 release note: 

    http://www-01.ibm.com/support/knowledgecenter/SSPT3X_3.0.0/com.ibm.swg.im.infosphere.biginsights.product.doc/doc/release_notes.html

    Tuong

    Hi Tuong. Thanks for your answer. I've tried to install BI EE 3.0 with HDFS and without adaptive MapReduce on a new clean VM to see if the problem is caused by GPFS. I installed the library libnuma (I think I didn't notice it during configuration before) as you suggested and everything worked! You definitely saved me some time. Now I try to install it with adaptive map reduce and GPFS. 

    I would like to use this opportunity and ask you (or others) one extra question about GPFS. I plan to install BI EE on 1 machine and than copy the file with VM. These copied VMs will be slaves. This works when using HDFS (or at least was in QE tutorial and worked for me). Is it possible to use the same procedure with GPFS?

  • 7PJN_Martin_Mašek
    7PJN_Martin_Mašek
    17 Posts

    Re: Installation BI EE 3.0 - Failed to execute hive commands GRANT

    ‏2014-08-06T19:05:20Z  

    Martin,

    Did you run a proper uninstall - uninstall.sh using your biadmin login?   These issue should be taken care of by the uninstaller.

    In order to remove of DB2 residual artifacts manually, you need to remove a couple of things:

    1. Remove the sqllib directory - for catalog user (on catalog node) as well as BigSQL userid on all other nodes

    2. On each node DB2 was installed (practically all node), remove all files under  /var/db2/

    3. Edit /etc/services file and remove any DB2 service port for catalog and BigSQL on all nodes.

        should look something like:

    db2c_catalog    xxxxx/tcp
    DB2_bigsql      xxxxx/tcp
    DB2_bigsql_END  xxxxxtcp
    db2c_bigsql     xxxxxtcp

     

    Hope this will get you moving again..

    Tuong
     

    I've run unistall.sh and it helped with problems except the DB2. I already deleted the VM so I cant try your steps but I definitely save your answer for the next time.

  • ShantanuMundkur
    ShantanuMundkur
    7 Posts

    Re: Installation BI EE 3.0 - Failed to execute hive commands GRANT

    ‏2014-08-08T18:31:27Z  

    Hi Tuong. Thanks for your answer. I've tried to install BI EE 3.0 with HDFS and without adaptive MapReduce on a new clean VM to see if the problem is caused by GPFS. I installed the library libnuma (I think I didn't notice it during configuration before) as you suggested and everything worked! You definitely saved me some time. Now I try to install it with adaptive map reduce and GPFS. 

    I would like to use this opportunity and ask you (or others) one extra question about GPFS. I plan to install BI EE on 1 machine and than copy the file with VM. These copied VMs will be slaves. This works when using HDFS (or at least was in QE tutorial and worked for me). Is it possible to use the same procedure with GPFS?

    I have not tried this myself but I highly doubt copying the VM file to create a slave would work with GPFS. In fact, to make it work even with HDFS, do you not need to update some configuration on the slave after copying ithe file? The equivalent (or at least what I see as equivalent) isn't trivially done with GPFS. So I would say it would not be possible to use the same procedure. Can you point to the specific QE tutorial so we have a better context of what you exactly you might have tried?

  • 7PJN_Martin_Mašek
    7PJN_Martin_Mašek
    17 Posts

    Re: Installation BI EE 3.0 - Failed to execute hive commands GRANT

    ‏2014-08-11T12:41:41Z  

    I have not tried this myself but I highly doubt copying the VM file to create a slave would work with GPFS. In fact, to make it work even with HDFS, do you not need to update some configuration on the slave after copying ithe file? The equivalent (or at least what I see as equivalent) isn't trivially done with GPFS. So I would say it would not be possible to use the same procedure. Can you point to the specific QE tutorial so we have a better context of what you exactly you might have tried?

    Hi. The video I was talking about is this one https://www.youtube.com/watch?v=gUal2orKVSY . I'm reading about installation and I think I will deploy 3 images into the cloud and then install BI without copying anything. But anyway. There has to be a way how to dynamically add slaves and configure what components I want on each slave isn't it?

    Updated on 2014-08-11T15:38:24Z at 2014-08-11T15:38:24Z by 7PJN_Martin_Mašek
  • ShantanuMundkur
    ShantanuMundkur
    7 Posts

    Re: Installation BI EE 3.0 - Failed to execute hive commands GRANT

    ‏2014-08-11T17:17:56Z  

    Hi. The video I was talking about is this one https://www.youtube.com/watch?v=gUal2orKVSY . I'm reading about installation and I think I will deploy 3 images into the cloud and then install BI without copying anything. But anyway. There has to be a way how to dynamically add slaves and configure what components I want on each slave isn't it?

    I understand your question better now.

    I see that the instructions in the video involves uninstalling BigInsights on the copy to enable the VM to be used as a node that can then be added to the cluster to achieve a multi-node cluster. So that approach should roughly work. I say "roughly" because

    1) The BigInsights QSE does not include GPFS

    2) If you somehow happened to have a similar VM with EE, then in addition to uninstalling BigInsights, you would also have to uninstall GPFS including cleaning up the node in a way that would not result it in it being detected as a node of an existing cluster or its disks as being used in another GPFS file system.

    When you add the node to the cluster, you are given the choice to add the GPFS component and you would typically always do this for nodes where you intend to run any file system-dependent services. Adding GPFS on a node includes installing RPMs and the subsequent steps to add node to the GPFS cluster and any specified disks to the file system. So you would need "clean" VMs for the additional nodes.

    3) Adding nodes in this manner will not given you an optimal setup especially starting with a single-node cluster. For instance, the add node functionality does not provide a way to specify that any of the added node's disks are to be used for metadata. This means you would end up with a configuration with metadata only on the master node and constrained to replication factor of 1. There are other configuration updates that would be desirable post-node-addition. Since this might not be obvious I thought I'd mention this. Unless you are familiar enough with GPFS configuration to carry out those updates post-install / node-addition, it would be better to run the BigInsights installation specifying all the nodes you want in the cluster.

    Updated on 2014-08-11T17:24:23Z at 2014-08-11T17:24:23Z by ShantanuMundkur
  • 7PJN_Martin_Mašek
    7PJN_Martin_Mašek
    17 Posts

    Re: Installation BI EE 3.0 - Failed to execute hive commands GRANT

    ‏2014-08-11T17:29:51Z  

    I understand your question better now.

    I see that the instructions in the video involves uninstalling BigInsights on the copy to enable the VM to be used as a node that can then be added to the cluster to achieve a multi-node cluster. So that approach should roughly work. I say "roughly" because

    1) The BigInsights QSE does not include GPFS

    2) If you somehow happened to have a similar VM with EE, then in addition to uninstalling BigInsights, you would also have to uninstall GPFS including cleaning up the node in a way that would not result it in it being detected as a node of an existing cluster or its disks as being used in another GPFS file system.

    When you add the node to the cluster, you are given the choice to add the GPFS component and you would typically always do this for nodes where you intend to run any file system-dependent services. Adding GPFS on a node includes installing RPMs and the subsequent steps to add node to the GPFS cluster and any specified disks to the file system. So you would need "clean" VMs for the additional nodes.

    3) Adding nodes in this manner will not given you an optimal setup especially starting with a single-node cluster. For instance, the add node functionality does not provide a way to specify that any of the added node's disks are to be used for metadata. This means you would end up with a configuration with metadata only on the master node and constrained to replication factor of 1. There are other configuration updates that would be desirable post-node-addition. Since this might not be obvious I thought I'd mention this. Unless you are familiar enough with GPFS configuration to carry out those updates post-install / node-addition, it would be better to run the BigInsights installation specifying all the nodes you want in the cluster.

    Thank you for response. You brought some light into the problem. I try to install BI with all machines presented during installation. It should be much easier.