Uploaded image for project: 'MariaDB Server'
  1. MariaDB Server
  2. MDEV-10128

Port MySQL 5.7 select count() optimization

Details

    Description

      Current handler::records() is not fully usable as InnoDB could roll back the statement on some cases and you may not continue after that, it should be int records(ha_rows *n_rows) and server should check as these errors will abort the current query:
      case HA_ERR_LOCK_DEADLOCK:
      case HA_ERR_LOCK_TABLE_FULL:
      case HA_ERR_LOCK_WAIT_TIMEOUT:
      case HA_ERR_QUERY_INTERRUPTED:

      see commit 40ec5373c0 on 5.7

      Attachments

        Issue Links

          relates to

          Task - A task that needs to be done. MDEV-11751 Merge new release of InnoDB MySQL 5.7.18 to 10.2

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

          New Feature - A new feature of the product, which has yet to be developed. MDEV-18188 Maintain persistent COUNT(*) in InnoDB

          • Major - Major loss of function.
          • Stalled

          Activity

            Ascending order - Click to sort in descending order
            marko Marko Mäkelä added a comment -

            Oracle removed this in MySQL 5.7.18, because it is hard-wired to use the clustered index.
            Sometimes using a secondary index could be a bad idea, because we do not have a hidden DB_TRX_ID column in the secondary indexes. For delete-marked records or for secondary index pages where PAGE_MAX_TRX_ID is too new, a lookup for a matching clustered index record must be performed (random lookups to the clustered index and optionally to undo log records). This is why it was hard-wired to use the clustered index in the first place.

            See also my old blog post that mentions InnoDB secondary indexes.

            marko Marko Mäkelä added a comment - Oracle removed this in MySQL 5.7.18 , because it is hard-wired to use the clustered index. Sometimes using a secondary index could be a bad idea, because we do not have a hidden DB_TRX_ID column in the secondary indexes. For delete-marked records or for secondary index pages where PAGE_MAX_TRX_ID is too new, a lookup for a matching clustered index record must be performed (random lookups to the clustered index and optionally to undo log records). This is why it was hard-wired to use the clustered index in the first place. See also my old blog post that mentions InnoDB secondary indexes.
            marko Marko Mäkelä added a comment -

            MDEV-11751 (Merge MySQL 5.7.18 to MariaDB 10.2) is removing the count optimization.

            marko Marko Mäkelä added a comment - MDEV-11751 (Merge MySQL 5.7.18 to MariaDB 10.2) is removing the count optimization.

            People

              marko Marko Mäkelä
              jplindst Jan Lindström (Inactive)
              Votes:
              0 Vote for this issue
              Watchers:
              3 Start watching this issue

              Dates

                Created:
                Updated:
                Resolved:

                Git Integration

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