Uploaded image for project: 'MariaDB Server'
  1. MariaDB Server
  2. MDEV-20133

Merge new release of InnoDB 5.7.27 to 10.2

Details

    Description

      MySQL 5.7.27 includes the following InnoDB changes:

      Not applicable to MariaDB Server

      Bug#28687608 EXCHANGE PARTITION GENERATING ERROR FOR ROW_FORMAT WHEN IT SHOULD NOT
      The code changes are related to the InnoDB native partitioning (ha_innopart) that was introduced in MySQL 5.7. At the very end of the extended test main.partition_exchange there is a test for the scenario, and MariaDB is correctly returning ER_PARTITION_EXCHANGE_DIFFERENT_OPTION. Note: the test assumes default_storage_engine='innodb', which does not hold in MariaDB. It also uses WITH VALIDATION and WITHOUT VALIDATION syntax, which is missing from MariaDB.

      Bug #29508001 MYSQL DEADLOCK AND BUGCHECK ON AARCH64 UNDER STRESS TEST
      This attempts to fix something that was already fixed in MariaDB in MDEV-7148, MDEV-10813, and MDEV-11296.

      Bug#28834208 row_log_table_apply_update*inplace_alter_table_impl*assertion failure:row0log.cc
      MariaDB Server 10.2.11, 10.3.3 and later refuse online rebuild if indexed virtual columns exist (MDEV-13795, MDEV-14332) because various problems exist with the logging.

      Bug#28104394 : INNODB 5.7 PRIMARY KEY SCAN LACK DATA
      This is one more fix for the "optimized range scan" series of regressions caused by a change that we decided not to merge to MariaDB in MDEV-11751.

      Not directly applicable to MariaDB

      Bug #25289359 DML/DDL ON LARGE FULLTEXT TABLES CAUSE SEMAPHORE TIMEOUTS AND ASSERTION/SUICIDE
      There is no test case, but there is debug instrumentation. We will investigate the scenario and track this change in MDEV-20126.

      Applicable to MariaDB Server

      Bug #29127203 VIRTUAL GENERATED COLUMN INDEX DATA INCONSISTENCY
      This appears to fix something with virtual columns that are connected to FOREIGN KEY constraints. There is no test case.

      BUG#29138644 PAGE CLEANER WILL SLEEP FOR LONG TIME IF CLOCK CHANGES and follow-up fix
      This will be handled in MDEV-14154. The functions ut_time_us(), ut_time_ms(), ut_time() must be replaced with the use of a a monotonic timer, such as my_interval_timer().

      Attachments

        Issue Links

          Activity

            The MDEV-14154 related changes covered almost all aspects of
            BUG#29138644 PAGE CLEANER WILL SLEEP FOR LONG TIME IF CLOCK CHANGES
            except this one:

            -	buf_pool->last_printout_time = ut_time();
            +	buf_pool->last_printout_time = ut_time_monotonic();
            

            This will affect time_elapsed in buf_stats_get_pool_info(), which in turn will affect the following:

            pool_info->page_made_young_rate
            pool_info->page_not_made_young_rate
            pool_info->pages_read_rate
            pool_info->pages_created_rate
            pool_info->pages_written_rate
            pool_info->pages_readahead_rnd_rate
            pool_info->pages_readahead_rate
            pool_info->pages_evicted_rate
            

            None of these affect the page_cleaner. They affect some diagnostic output, such as SHOW ENGINE INNODB STATUS and INFORMATION_SCHEMA.INNODB_BUFFER_POOL_STATS.
            Because time(NULL) may have significantly smaller overhead than a monotonic timer, it is best to avoid replacing that.

            marko Marko Mäkelä added a comment - The MDEV-14154 related changes covered almost all aspects of BUG#29138644 PAGE CLEANER WILL SLEEP FOR LONG TIME IF CLOCK CHANGES except this one: - buf_pool->last_printout_time = ut_time(); + buf_pool->last_printout_time = ut_time_monotonic(); This will affect time_elapsed in buf_stats_get_pool_info() , which in turn will affect the following: pool_info->page_made_young_rate pool_info->page_not_made_young_rate pool_info->pages_read_rate pool_info->pages_created_rate pool_info->pages_written_rate pool_info->pages_readahead_rnd_rate pool_info->pages_readahead_rate pool_info->pages_evicted_rate None of these affect the page_cleaner . They affect some diagnostic output, such as SHOW ENGINE INNODB STATUS and INFORMATION_SCHEMA.INNODB_BUFFER_POOL_STATS . Because time(NULL) may have significantly smaller overhead than a monotonic timer, it is best to avoid replacing that.

            People

              marko Marko Mäkelä
              marko Marko Mäkelä
              Votes:
              0 Vote for this issue
              Watchers:
              1 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.