Pig
  1. Pig
  2. PIG-2711

e2e harness: cache benchmark results between test runs

    Details

    • Type: Improvement Improvement
    • Status: Closed
    • Priority: Major Major
    • Resolution: Fixed
    • Affects Version/s: None
    • Fix Version/s: 0.11, 0.10.1
    • Component/s: e2e harness
    • Labels:
      None
    • Hadoop Flags:
      Reviewed

      Description

      It will be helpful to be able to store the benchmark results that are computed with "old pig" and reuse in subsequent runs.

      1. PIG-2711.patch
        3 kB
        Thomas Weise
      2. PIG-2711-2.patch
        0.5 kB
        Daniel Dai

        Activity

        Hide
        Thomas Weise added a comment -

        For one this will speed up subsequent runs as we eliminate redundant pig runs (on some clusters running the harness at present takes up to 24 hours with local and mapred mode and additional internal tests).

        This feature can also be used to compare pig and hadoop versions based on archived benchmarks. While the harness currently allows to configure old and new pig and hadoop versions, it would be much simpler to configure the benchmark than recomputing it.

        Show
        Thomas Weise added a comment - For one this will speed up subsequent runs as we eliminate redundant pig runs (on some clusters running the harness at present takes up to 24 hours with local and mapred mode and additional internal tests). This feature can also be used to compare pig and hadoop versions based on archived benchmarks. While the harness currently allows to configure old and new pig and hadoop versions, it would be much simpler to configure the benchmark than recomputing it.
        Hide
        Daniel Dai added a comment -

        Patch committed to trunk. Thanks Thomas!

        Show
        Daniel Dai added a comment - Patch committed to trunk. Thanks Thomas!
        Hide
        Daniel Dai added a comment -

        Also commit to 0.10 branch as per Thomas' request.

        Show
        Daniel Dai added a comment - Also commit to 0.10 branch as per Thomas' request.
        Hide
        Jie Li added a comment -

        Do we need to update the documentation for setting up the e2e environment? I couldn't get it work due to the following error:

        (MR test job running succesfully...)
        ERROR TestDriver::run at : 470 Failed to run test FilterBoolean_1 <Unable to open file $

        {PH_BENCHMARK_CACHE_PATH}

        /FilterBoolean_1_benchmark.pig to write pig script, No such file or directory
        >

        Even I passed -DPH_BENCHMARK_CACHE_PATH=some_path, I still got the error:

        ERROR TestDriver::run at : 470 Failed to run test FilterBoolean_1 <Unable to open file /Users/JieLi/svn/pig/pig_benchmark_cache/FilterBoolean_1_benchmark.pig to write pig script, No such file or directory

        Show
        Jie Li added a comment - Do we need to update the documentation for setting up the e2e environment? I couldn't get it work due to the following error: (MR test job running succesfully...) ERROR TestDriver::run at : 470 Failed to run test FilterBoolean_1 <Unable to open file $ {PH_BENCHMARK_CACHE_PATH} /FilterBoolean_1_benchmark.pig to write pig script, No such file or directory > Even I passed -DPH_BENCHMARK_CACHE_PATH=some_path, I still got the error: ERROR TestDriver::run at : 470 Failed to run test FilterBoolean_1 <Unable to open file /Users/JieLi/svn/pig/pig_benchmark_cache/FilterBoolean_1_benchmark.pig to write pig script, No such file or directory
        Hide
        Thomas Weise added a comment -

        What you specify needs to be an existing path. Did you create /Users/JieLi/svn/pig/pig_benchmark_cache ?

        Show
        Thomas Weise added a comment - What you specify needs to be an existing path. Did you create /Users/JieLi/svn/pig/pig_benchmark_cache ?
        Hide
        Thomas Weise added a comment -

        Also, this was meant to be optional (see 0.10 branch). The build.xml in trunk may have additional modifications that don't set PH_BENCHMARK_CACHE_PATH empty by default.

        Show
        Thomas Weise added a comment - Also, this was meant to be optional (see 0.10 branch). The build.xml in trunk may have additional modifications that don't set PH_BENCHMARK_CACHE_PATH empty by default.
        Hide
        Jie Li added a comment -

        Thanks Thomas! That works. How about adding this to https://cwiki.apache.org/confluence/display/PIG/HowToTest?

        Show
        Jie Li added a comment - Thanks Thomas! That works. How about adding this to https://cwiki.apache.org/confluence/display/PIG/HowToTest?
        Hide
        Daniel Dai added a comment -

        Yes, it mean to be optional. Attach the patch for fix. Can you try it Jie?

        Show
        Daniel Dai added a comment - Yes, it mean to be optional. Attach the patch for fix. Can you try it Jie?
        Hide
        Jie Li added a comment -

        Sure let me try it

        Show
        Jie Li added a comment - Sure let me try it
        Hide
        Daniel Dai added a comment -

        Tested. Commit PIG-2711-2.patch to 0.10/trunk.

        Show
        Daniel Dai added a comment - Tested. Commit PIG-2711 -2.patch to 0.10/trunk.

          People

          • Assignee:
            Thomas Weise
            Reporter:
            Thomas Weise
          • Votes:
            0 Vote for this issue
            Watchers:
            3 Start watching this issue

            Dates

            • Created:
              Updated:
              Resolved:

              Development