You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@hive.apache.org by Jay Booth <ja...@gmail.com> on 2010/02/09 22:24:14 UTC

Re: Hive uses hadoop version information which unfortunately is unknown

Hey Ed, if you have the option of recompiling hadoop without updating, you
can simply run "src/saveVersion.sh 0.18.3 ; ant bin-package", and that will
tag your binaries with the 0.18.3 label.

If recompiling's not an option..  sorry.  This one's bitten us a couple
times when rolling out new hadoop versions as well.

On Tue, Feb 9, 2010 at 2:12 PM, Edward Capriolo <ed...@gmail.com>wrote:

> I would like to run hive release, unfortunately I have an older 18.3
> hadoop that was build without version information.
>
> I am going to check out the 4.1 release and hack at the shell scripts
> and the relevant code in hive. Does someone have the output of 'hadoop
> version' 18.3 on 18.3, or a better way to handle my problem?
> (I know the correct answer is update hadoop but currently I am not
> going to follow this course)
>
> Thanks
> Edward
>

Re: Hive uses hadoop version information which unfortunately is unknown

Posted by Edward Capriolo <ed...@gmail.com>.
On Tue, Feb 9, 2010 at 4:24 PM, Jay Booth <ja...@gmail.com> wrote:
> Hey Ed, if you have the option of recompiling hadoop without updating, you
> can simply run "src/saveVersion.sh 0.18.3 ; ant bin-package", and that will
> tag your binaries with the 0.18.3 label.
> If recompiling's not an option..  sorry.  This one's bitten us a couple
> times when rolling out new hadoop versions as well.
>
> On Tue, Feb 9, 2010 at 2:12 PM, Edward Capriolo <ed...@gmail.com>
> wrote:
>>
>> I would like to run hive release, unfortunately I have an older 18.3
>> hadoop that was build without version information.
>>
>> I am going to check out the 4.1 release and hack at the shell scripts
>> and the relevant code in hive. Does someone have the output of 'hadoop
>> version' 18.3 on 18.3, or a better way to handle my problem?
>> (I know the correct answer is update hadoop but currently I am not
>> going to follow this course)
>>
>> Thanks
>> Edward
>
>

Jay,
Thanks. Recompile on hadoop is not an option. 18.3, Its a can of worms :)

Here is what I did, to bypass the unknown. This is bad and breaks the
shim capabilities of hive.
[ecapriolo@hadoopdata10 hive-4.1-rc2]$ svn diff
Index: shims/src/common/java/org/apache/hadoop/hive/shims/ShimLoader.java
===================================================================
--- shims/src/common/java/org/apache/hadoop/hive/shims/ShimLoader.java
 (revision 908160)
+++ shims/src/common/java/org/apache/hadoop/hive/shims/ShimLoader.java
 (working copy)
@@ -94,6 +94,8 @@
    * (e.g "0.18" or "0.20")
    */
   private static String getMajorVersion() {
+    return "0.18";
+    /*
     String vers = VersionInfo.getVersion();

     String parts[] = vers.split("\\.");
@@ -102,5 +104,6 @@
                                  " (expected A.B.* format)");
     }
     return parts[0] + "." + parts[1];
+    */
   }
 }
Index: bin/ext/cli.sh
===================================================================
--- bin/ext/cli.sh      (revision 908160)
+++ bin/ext/cli.sh      (working copy)
@@ -15,7 +15,7 @@
   fi

   version=$($HADOOP version | awk '{if (NR == 1) {print $2;}}');
-
+  version="0.18.3"
   # Save the regex to a var to workaround quoting incompatabilities
   # between Bash 3.1 and 3.2
   version_re="^([[:digit:]]+)\.([[:digit:]]+)(\.([[:digit:]]+))?.*$"
[ecapriolo@nyhadoopdata10 hive-4.1-rc2]$