Details

    • Bug
    • Status: Closed (View Workflow)
    • Major
    • Resolution: Fixed
    • 10.2.15
    • 10.2.32, 10.3.23, 10.4.13
    • Protocol
    • None
    • Windows Server 2008 R2.
      MySQL Connector Net 6.3.5 (MySQL Connector Net 6.9.11)
      .net 4.5.

    Description

      После установки mariaDB 10.2.15 начала появляться ошибка:
      MySql.Data.MySqlClient.MySqlException: Out of sync with server

      Если сразу после ошибки дать повторный запрос - ошибки не будет. В интете не нашел ничего похожего на мой случай. Скрипты разные.
      Ошибка не зависит от нагрузки (количества соединений). Она проявляется и ночью, когда нагрузка минимальная, и днем.
      Допускаю, что ошибка не связана с версией, но, вроде, других изменений, которые могли привести к данной ошибке не было.

      Attachments

        1. my.ini
          2 kB
          Sergii Volikov
        2. MySqlErr.png
          19 kB
          Sergii Volikov
        3. MySqlErrlog.pcap
          37 kB
          Vladislav Vaintroub

        Issue Links

          Activity

            Moshe L Moshe L added a comment - - edited

            Thank you for all work.

            I want to do it. yes.
            the problem is that the problem is not deterministic. on the same code it failed about 1/1000~ of the runs, so for repeat it I running a loop until it throwing the error.

            the strage thing: I was saw the errors on my production (3 diffrent machines) but non on my development system.
            I was upgraded the server to 10.4.10 some months ago, and after this upgrade (before it was 10.3.8) I see less error messages.

            btw, I was saw a "better" error rate on this query, on a simple loop:

            {{
            /*CREATE TABLE `sites_addresses` (
            `sa_id` int(11) NOT NULL AUTO_INCREMENT,
            `sa_site` int(11) NOT NULL,
            `sa_address` int(11) NOT NULL,
            `sa_visible` tinyint(1) NOT NULL DEFAULT 1,
            PRIMARY KEY (`sa_id`),
            UNIQUE KEY `site_address` (`sa_site`,`sa_address`),
            KEY `sad_visible` (`sa_visible`,`sa_id`,`sa_site`,`sa_address`),
            KEY `filter_visible` (`sa_address`,`sa_site`,`sa_visible`)
            ) ENGINE=InnoDB AUTO_INCREMENT=897110 DEFAULT CHARSET=utf8mb4;
            */
            string sql = @"
            insert into sites_addresses set sa_site=1218006, sa_visible=1, sa_address=66968 on duplicate key update sa_site=1218006, sa_visible=1, sa_address=66968; select sa_id from sites_addresses where sa_address=66968 and sa_site=1218006";
            }}

            the specific error message is diffrent:

            `Packet received out-of-order. Expected 1; got 115., /Company/PostSite`

            on the old testcase it was `got 84`.

            Moshe L Moshe L added a comment - - edited Thank you for all work. I want to do it. yes. the problem is that the problem is not deterministic . on the same code it failed about 1/1000~ of the runs, so for repeat it I running a loop until it throwing the error. the strage thing: I was saw the errors on my production (3 diffrent machines) but non on my development system. I was upgraded the server to 10.4.10 some months ago, and after this upgrade (before it was 10.3.8) I see less error messages. btw, I was saw a "better" error rate on this query, on a simple loop: {{ /*CREATE TABLE `sites_addresses` ( `sa_id` int(11) NOT NULL AUTO_INCREMENT, `sa_site` int(11) NOT NULL, `sa_address` int(11) NOT NULL, `sa_visible` tinyint(1) NOT NULL DEFAULT 1, PRIMARY KEY (`sa_id`), UNIQUE KEY `site_address` (`sa_site`,`sa_address`), KEY `sad_visible` (`sa_visible`,`sa_id`,`sa_site`,`sa_address`), KEY `filter_visible` (`sa_address`,`sa_site`,`sa_visible`) ) ENGINE=InnoDB AUTO_INCREMENT=897110 DEFAULT CHARSET=utf8mb4; */ string sql = @" insert into sites_addresses set sa_site=1218006, sa_visible=1, sa_address=66968 on duplicate key update sa_site=1218006, sa_visible=1, sa_address=66968; select sa_id from sites_addresses where sa_address=66968 and sa_site=1218006"; }} the specific error message is diffrent: `Packet received out-of-order. Expected 1; got 115., /Company/PostSite` on the old testcase it was `got 84`.

            https://1drv.ms/u/s!AqT4tBcElQz_bgxYQBf6J2ykN_A?e=h6uH6N

            I build a ZIP package for the "fix by suspicion" ( https://github.com/mariadb/server/commit/64437d040bfdb61c92efd33301a676ca02f342c8 ).
            Moshe L, I would appreciate if you can give it a test? It is on top of current 10.4, with the patch I mention.

            I tend to strongly suspect the COM_MULTI code that came in 10.2 for the breakage. that line that I removed was protocol-relevant addition that came with that code, and this is in semicolon batching. And it kinda rhymes well with "concatenated packets" that we have seen in the network trace

            sanja, FYI.

            wlad Vladislav Vaintroub added a comment - https://1drv.ms/u/s!AqT4tBcElQz_bgxYQBf6J2ykN_A?e=h6uH6N I build a ZIP package for the "fix by suspicion" ( https://github.com/mariadb/server/commit/64437d040bfdb61c92efd33301a676ca02f342c8 ). Moshe L , I would appreciate if you can give it a test? It is on top of current 10.4, with the patch I mention. I tend to strongly suspect the COM_MULTI code that came in 10.2 for the breakage. that line that I removed was protocol-relevant addition that came with that code, and this is in semicolon batching. And it kinda rhymes well with "concatenated packets" that we have seen in the network trace sanja , FYI.
            Moshe L Moshe L added a comment -

            my test:

            extract your zip file
            backup all
            stop mysql service
            run from commandline the exact data/my.ini/settings but on the exe from your zip, on production server

            run the testcase

            The testcase success. I will wait a day to see if the random problem disappears.

            Moshe L Moshe L added a comment - my test: extract your zip file backup all stop mysql service run from commandline the exact data/my.ini/settings but on the exe from your zip, on production server run the testcase The testcase success. I will wait a day to see if the random problem disappears.

            . Thank you Moshe L. Please keep us informed.

            wlad Vladislav Vaintroub added a comment - . Thank you Moshe L . Please keep us informed.
            Moshe L Moshe L added a comment - - edited

            4 days of quiet.

            you can mark this bug as fixed, I think.

            Moshe L Moshe L added a comment - - edited 4 days of quiet. you can mark this bug as fixed, I think.

            People

              sanja Oleksandr Byelkin
              Sergey_W Sergii Volikov
              Votes:
              0 Vote for this issue
              Watchers:
              10 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.