Uploaded image for project: 'MariaDB Server'
  1. MariaDB Server
  2. MDEV-11565

possible bug? #innodb-encryption-threads

Details

    Description

      hi team

      I'm having a peculiar situation, a problem here with my config, which config is pretty galera-vanilla.
      Namely, if I set these:

      plugin-load-add = file_key_management.so
      		 
      file-key-management
      		 
      file-key-management-filename = /var/lib/mysql/.keys.txt
      		 
      innodb_file_per_table = 1
      		 
      innodb-encrypt-log
      		 
      innodb-encryption-threads = 4     <= # this ! one here

      I get constant load on four cores, it would fluctuate just under %400 CPU (24*7). I'd assume that innodb-encryption-threads value causes it, and whatever the value might be it will be reflected in CPU usage.
      I wonder if you too see it.
      I my case I think it does something more to one system, it crashes the bloody thing to the extent where cold reboot is needed.

      MariaDB-client-10.1.19-1.el7.centos.x86_64
      		 
      MariaDB-shared-10.1.19-1.el7.centos.x86_64
      		 
      MariaDB-common-10.1.19-1.el7.centos.x86_64
      		 
      MariaDB-server-10.1.19-1.el7.centos.x86_64

      best,
      L

      Attachments

        Activity

          Ascending order - Click to sort in descending order
          elenst Elena Stepanova added a comment -

          I am not getting this, but we've had similar complaints before, so I'm sure it can happen. There is some explanation in MDEV-11581, please take a look, maybe it will help.

          elenst Elena Stepanova added a comment - I am not getting this, but we've had similar complaints before, so I'm sure it can happen. There is some explanation in MDEV-11581 , please take a look, maybe it will help.
          jplindst Jan Lindström (Inactive) added a comment -

          Naturally, if you set up innodb-encryption-threads = 4, these threads will periodically check all tablespaces if they need (1) key rotation, (2) transform tablespace from unencrypted to encrypted, (3) transform tablespace from encrypted to unencrypted and (4) for scrubbing (physically zero delete marked rows). If you do not want or need any of this you may very well set innodb-encryption-threads = 0 and all old and new tables you need to encrypt will be encrypted if create table contains ENCRYPTED=yes. If you use ENCRYPTED=default or do not provide ENCRYPTED option but still expect you tables to be encrypted you need to set these threads.

          jplindst Jan Lindström (Inactive) added a comment - Naturally, if you set up innodb-encryption-threads = 4, these threads will periodically check all tablespaces if they need (1) key rotation, (2) transform tablespace from unencrypted to encrypted, (3) transform tablespace from encrypted to unencrypted and (4) for scrubbing (physically zero delete marked rows). If you do not want or need any of this you may very well set innodb-encryption-threads = 0 and all old and new tables you need to encrypt will be encrypted if create table contains ENCRYPTED=yes. If you use ENCRYPTED=default or do not provide ENCRYPTED option but still expect you tables to be encrypted you need to set these threads.
          jplindst Jan Lindström (Inactive) added a comment -

          Works as designed, consider used configuration and if needed correct the number of background encryption threads.

          jplindst Jan Lindström (Inactive) added a comment - Works as designed, consider used configuration and if needed correct the number of background encryption threads.
          lejeczek none now added a comment -

          but the thing which makes it peculiar in my case and which made me report it here is - it happens to mariadb with databases where there is nothing going on, nothing is happening. No activities, no queries, but that constant %400 cpu load remains. Is this still normal, expected?
          regards

          lejeczek none now added a comment - but the thing which makes it peculiar in my case and which made me report it here is - it happens to mariadb with databases where there is nothing going on, nothing is happening. No activities, no queries, but that constant %400 cpu load remains. Is this still normal, expected? regards
          jplindst Jan Lindström (Inactive) added a comment -

          Yes normal, expected. Those threads will do their work (i.e. check every tablespace) even if nothing else is done.

          jplindst Jan Lindström (Inactive) added a comment - Yes normal, expected. Those threads will do their work (i.e. check every tablespace) even if nothing else is done.

          People

            jplindst Jan Lindström (Inactive)
            lejeczek none now
            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.