Cache non-exist timeseries metadata to reduce IO #16606
                
     Open
            
            
          
  Add this suggestion to a batch that can be applied as a single commit.
  This suggestion is invalid because no changes were made to the code.
  Suggestions cannot be applied while the pull request is closed.
  Suggestions cannot be applied while viewing a subset of changes.
  Only one suggestion per line can be applied in a batch.
  Add this suggestion to a batch that can be applied as a single commit.
  Applying suggestions on deleted lines is not supported.
  You must change the existing code in this line in order to create a valid suggestion.
  Outdated suggestions cannot be applied.
  This suggestion has been applied or marked resolved.
  Suggestions cannot be applied from pending reviews.
  Suggestions cannot be applied on multi-line comments.
  Suggestions cannot be applied while the pull request is queued to merge.
  Suggestion cannot be applied right now. Please check back later.
  
    
  
    


When repeatedly querying a timeseries that is not in a TsFile and the bloomfilter fails to filter it, each query will result in a necessary IO.
Since TsFile is immutable, we can cache a placeholder to indicate the non-existence.
In a test, 10000 series were written into a TsFile, then 20000 series (only 10000 exist) were queried from the file.
The result shows a significant reduction in query time.
For 0.25M timeseries
For 1M timeseries