Uploaded image for project: 'MariaDB ColumnStore'
  1. MariaDB ColumnStore
  2. MCOL-5460

concurrent CTE operation excessive mutex

Details

    • Bug
    • Status: Stalled (View Workflow)
    • Critical
    • Resolution: Unresolved
    • 23.02.1
    • 23.10
    • None
    • 2023-8, 2023-10, 2023-11, 2023-12, 2024-1, 2024-2

    Attachments

      2. CTE1-23_10_0.png
        127 kB
        Roman
      3. CTE1-3-patches.png
        104 kB
        Roman
      4. CTE2-23_10_0.png
        135 kB
        Roman
      5. CTE2-3-patches.png
        120 kB
        Roman
      6. CTE3-23_10_0.png
        99 kB
        Roman
      7. CTE3-3-patches_2.png
        104 kB
        Roman
      8. CTE3-3-patches.png
        97 kB
        Roman
      9. screenshot-1.png
        174 kB
        Allen Herrera
      10. screenshot-2.png
        182 kB
        Allen Herrera
      11. screenshot-3.png
        202 kB
        Allen Herrera
      12. screenshot-4.png
        196 kB
        Allen Herrera

      Issue Links

        includes

        Bug - A problem which impairs or prevents the functions of the product. MCOL-5472 CPU wastes time in wait and sys b/c of urandom access in RowStorage.

        • Blocker - Blocks development and/or testing work, production could not run.
        • Closed

        New Feature - A new feature of the product, which has yet to be developed. MCOL-5473 Mutex lock contention doing VSS lookup loading a block

        • Critical - Crashes, loss of data, severe memory leak.
        • Stalled

        New Feature - A new feature of the product, which has yet to be developed. MCOL-5474 Mutex lock contention doing block cache lookup when loading a block

        • Critical - Crashes, loss of data, severe memory leak.
        • Stalled

        Bug - A problem which impairs or prevents the functions of the product. MCOL-5601 Config::makeConfig bottlenecks on a mutex contributing to CPU burned in sys

        • Major - Major loss of function.
        • Closed

        Activity

          Ascending order - Click to sort in descending order
          drrtuy Roman added a comment -

          bbancroft CSV is OK.

          drrtuy Roman added a comment - bbancroft CSV is OK.
          alexey.vorovich alexey vorovich (Inactive) added a comment -

          leonid.fedorov we need to connect PR to this ticket

          alexey.vorovich alexey vorovich (Inactive) added a comment - leonid.fedorov we need to connect PR to this ticket
          alexey.vorovich alexey vorovich (Inactive) added a comment -

          allen.herrera So did we recover the overall performance of concurrent queries enough ?

          alexey.vorovich alexey vorovich (Inactive) added a comment - allen.herrera So did we recover the overall performance of concurrent queries enough ?
          alexey.vorovich alexey vorovich (Inactive) added a comment -

          The key here is to estimate what regression in timing (or improvements) we have after the recent fixes as compared to 611

          alexey.vorovich alexey vorovich (Inactive) added a comment - The key here is to estimate what regression in timing (or improvements) we have after the recent fixes as compared to 611
          drrtuy Roman added a comment -

          I need an access to the machines to see where is the bottleneck.

          drrtuy Roman added a comment - I need an access to the machines to see where is the bottleneck.
          leonid.fedorov Leonid Fedorov added a comment -

          allen.herrera when you are referring 23.08.x release, that doesn't exist. Are you mentioning out last 23.02.2 package or 22.08.8? There is significant difference here

          leonid.fedorov Leonid Fedorov added a comment - allen.herrera when you are referring 23.08.x release, that doesn't exist. Are you mentioning out last 23.02.2 package or 22.08.8? There is significant difference here
          drrtuy Roman added a comment -

          Added bunch of screens with sysbench results to compare 23.10.0 and the build that has 3 patches out of 4.
          CTE3 didn't get any better that means there was no mutex contention in the first place.

          drrtuy Roman added a comment - Added bunch of screens with sysbench results to compare 23.10.0 and the build that has 3 patches out of 4. CTE3 didn't get any better that means there was no mutex contention in the first place.

          People

            drrtuy Roman
            alexey.vorovich alexey vorovich (Inactive)
            Votes:
            1 Vote for this issue
            Watchers:
            9 Start watching this issue

            Dates

              Created:
              Updated:

              Git Integration

                Error rendering 'com.xiplink.jira.git.jira_git_plugin:git-issue-webpanel'. Please contact your Jira administrators.