Uploaded image for project: 'Couchbase Server'
  1. Couchbase Server
  2. MB-47200

Provide detailed index scan plan/report

    XMLWordPrintable

Details

    • Improvement
    • Status: Open
    • Major
    • Resolution: Unresolved
    • None
    • Morpheus, Neo.next
    • secondary-index
    • None

    Description

      Query service provides users with a query execution plan which includes details like time spent in each phase of query execution. It includes total time spent in index scan. It will be useful if more timing details for index scan are provided as a part of the report.

      Possible list of metrics reported can be a follows
      1. Total scan duration on indexer side
      2. Total scan wait duration on indexer side - including/excluding fetchseqnos.
      3. Total scan duration on query client side
      4. Total scan wait duration on query client side - including backfill file reading
      5. Total disk read duration on indexer side
      6. Total duration of scatter/gather per partitioned index on indexer as well as on client side.
      7. Total scan duration for group by and aggregate phase.
      etc.

      This will require protocol change with query service, between indexer and query client, and with index storage.

      Attachments

        No reviews matched the request. Check your Options in the drop-down menu of this sections header.

        Activity

          Some more useful stats for query profiling:

          1. cache miss ratio 
          2. num rows scanned (useful for aggregate scans/scans where rows returned is much less than scanned)
          3. index instance which served the scan(useful in cases where one node is slower than others)
          4. partition elimination information(did the scan require scatter gather for all partitions or partition elimination worked?)

           

          deepkaran.salooja Deepkaran Salooja added a comment - Some more useful stats for query profiling: cache miss ratio  num rows scanned (useful for aggregate scans/scans where rows returned is much less than scanned) index instance which served the scan(useful in cases where one node is slower than others) partition elimination information(did the scan require scatter gather for all partitions or partition elimination worked?)  

          People

            yogendra.acharya Yogendra Acharya (Inactive)
            amit.kulkarni Amit Kulkarni
            Votes:
            0 Vote for this issue
            Watchers:
            4 Start watching this issue

            Dates

              Created:
              Updated:

              Gerrit Reviews

                There are no open Gerrit changes

                PagerDuty