Running wikimediumall after wikimedium10k always fails

[jainankitk]

Following the README guide instructions of running wikimedium10k before wikimediumall always runs into below exception.

  run vmstat: /usr/bin/vmstat --active --wide --timestamp --unit M 1 > /local/home/akjain/ws/logs/baseline_vs_patch.        vmstat.log 2>&1 &                                                                                                       
  Traceback (most recent call last):                                                                                      
    File "/local/home/akjain/ws/util/src/python/localrun.py", line 85, in <module>                                        
      comp.benchmark("baseline_vs_patch")                                                                                 
    File "/local/home/akjain/ws/util/src/python/competition.py", line 547, in benchmark                                   
      searchBench.run(                                                                                                    
    File "/local/home/akjain/ws/util/src/python/searchBench.py", line 69, in run                                          
      r.makeIndex(id, c.index, doCharts)                                                                                  
    File "/local/home/akjain/ws/util/src/python/benchUtil.py", line 1059, in makeIndex                                    
      run(cmd, fullLogFile, vmstatLogFile=vmstatLogFile, topLogFile=topLogFile)                                           
    File "/local/home/akjain/ws/util/src/python/benchUtil.py", line 873, in run                                           
      topProcess = ps_head.PSTopN(10, topLogFile)                                                                         
    File "/local/home/akjain/ws/util/src/python/ps_head.py", line 39, in __init__                                         
      raise RuntimeError(f"please remove log file {log_file_name} first")                                                 
  RuntimeError: please remove log file /local/home/akjain/ws/logs/baseline_vs_patch.top.log first

IMO, we should do one of below things:

• Create run specific log directory for each run
• Delete log directory before starting the run. Probably, never a good idea to silently delete bunch of logs
• Update README instructions to delete the log directory before moving to wikimediumall

If feasible, my preference is for the first option.

[mikemccand]

Good fresh eyes catch @jainankitk -- thank you! +1 for that first option. Though, we will leak log directories over time ... I guess that's the lesser evil here?

[jainankitk]

Though, we will leak log directories over time ... I guess that's the lesser evil here?

I think so, since they will eventually need to clear those logs (once in few weeks/month), which needs to be done even today!

[prudhvigodithi]

Seen the same error

run vmstat: /usr/bin/vmstat --active --wide --timestamp --unit M 1 > /home/ec2-user/lucene/benchmarks/logs/baseline_vs_patch.vmstat.log 2>&1 &
Traceback (most recent call last):
  File "/home/ec2-user/lucene/benchmarks/util/src/python/localrun.py", line 85, in <module>
    comp.benchmark("baseline_vs_patch")
  File "/home/ec2-user/lucene/benchmarks/util/src/python/competition.py", line 551, in benchmark
    searchBench.run(
  File "/home/ec2-user/lucene/benchmarks/util/src/python/searchBench.py", line 69, in run
    r.makeIndex(id, c.index, doCharts)
  File "/home/ec2-user/lucene/benchmarks/util/src/python/benchUtil.py", line 1061, in makeIndex
    run(cmd, fullLogFile, vmstatLogFile=vmstatLogFile, topLogFile=topLogFile)
  File "/home/ec2-user/lucene/benchmarks/util/src/python/benchUtil.py", line 875, in run
    topProcess = ps_head.PSTopN(10, topLogFile)
  File "/home/ec2-user/lucene/benchmarks/util/src/python/ps_head.py", line 39, in __init__
    raise RuntimeError(f"please remove log file {log_file_name} first")
RuntimeError: please remove log file /home/ec2-user/lucene/benchmarks/logs/baseline_vs_patch.top.log first

[prudhvigodithi]

Created a PR to solve this #489, please check if this make sense.