Uploaded image for project: 'MariaDB Server'
  1. MariaDB Server
  2. MDEV-7139

Sporadic failure in innodb.innodb_corrupt_bit on P8

    Details

      Description

      http://buildbot.askmonty.org/buildbot/builders/bintar-rhel7-p8-debug/builds/12/steps/test/logs/stdio

      innodb.innodb_corrupt_bit 'innodb_plugin' w1 [ fail ]
              Test ended at 2014-11-19 05:33:19
       
      CURRENT_TEST: innodb.innodb_corrupt_bit
      mysqltest: At line 74: query 'INSERT INTO corrupt_bit_test_ā VALUES(13000,'x',1,1)' failed: 1062: Duplicate entry '1-x' for key 'idxē'
       
      The result from queries just before the failure was:
      < snip >
      @@unique_checks
      0
      select @@innodb_change_buffering_debug;
      @@innodb_change_buffering_debug
      1
      INSERT INTO corrupt_bit_test_ā SELECT 0,b,c+1,z+1 FROM corrupt_bit_test_ā;
      INSERT INTO corrupt_bit_test_ā SELECT 0,b,c+10,z+10 FROM corrupt_bit_test_ā;
      INSERT INTO corrupt_bit_test_ā SELECT 0,b,c+20,z+20 FROM corrupt_bit_test_ā;
      INSERT INTO corrupt_bit_test_ā SELECT 0,b,c+50,z+50 FROM corrupt_bit_test_ā;
      INSERT INTO corrupt_bit_test_ā SELECT 0,b,c+100,z+100 FROM corrupt_bit_test_ā;
      INSERT INTO corrupt_bit_test_ā SELECT 0,b,c+200,z+200 FROM corrupt_bit_test_ā;
      INSERT INTO corrupt_bit_test_ā SELECT 0,b,c+400,z+400 FROM corrupt_bit_test_ā;
      INSERT INTO corrupt_bit_test_ā SELECT 0,b,c+800,z+800 FROM corrupt_bit_test_ā;
      INSERT INTO corrupt_bit_test_ā SELECT 0,b,c+1600,z+1600 FROM corrupt_bit_test_ā;
      INSERT INTO corrupt_bit_test_ā SELECT 0,b,c+4000,z+4000 FROM corrupt_bit_test_ā;
      select count(*) from corrupt_bit_test_ā;
      count(*)
      1024
      CREATE INDEX idx3 ON corrupt_bit_test_ā(b, c);
      INSERT INTO corrupt_bit_test_ā VALUES(13000,'x',1,1);
       
      More results from queries before failure can be found in /home/buildbot/maria-slave/power8-vlp03-bintar-debug/build/mysql-test/var/1/log/innodb_corrupt_bit.log

      So far it happened twice, one time above on 10.0 / bintar-rhel7-p8-debug, and another one on 5.5 / bintar-trusty-p8-debug. Apparently there are no logs from the other one, but it can be seen in Cross Reference report:

      [Build Details]	5.5	4339	bintar-trusty-p8-debug	2014-10-31 15:00:29	172	nm	Normal run, no --ps-protocol	innodb.innodb_corrupt_bit	xtradb	
      innodb.innodb_corrupt_bit 'xtradb'       w4 [ fail ]
              Test ended at 2014-10-31 09:47:07
       
      CURRENT_TEST: innodb.innodb_corrupt_bit
      mysqltest: At line 72: query 'INSERT INTO corrupt_bit_test_ā VALUES(13000,'x',1,1)' failed: 1062: Duplicate entry '1-x' for key 'idxā'
       
      The result from queries just before the failure was:
      < snip >
      @@unique_checks
      0
      select @@innodb_change_buffering_debug;
      @@innodb_change_buffering_debug
      1
      INSERT INTO corrupt_bit_test_ā SELECT 0,b,c+1,z+1 FROM corrupt_bit_test_ā;
      INSERT INTO corrupt_bit_test_ā SELECT 0,b,c+10,z+10 FROM corrupt_bit_test_ā;
      INSERT INTO corrupt_bit_test_ā SELECT 0,b,c+20,z+20 FROM corrupt_bit_test_ā;
      INSERT INTO corrupt_bit_test_ā SELECT 0,b,c+50,z+50 FROM corrupt_bit_test_ā;
      INSERT INTO corrupt_bit_test_ā SELECT 0,b,c+100,z+100 FROM corrupt_bit_test_ā;
      INSERT INTO corrupt_bit_test_ā SELECT 0,b,c+200,z+200 FROM corrupt_bit_test_ā;
      INSERT INTO corrupt_bit_test_ā SELECT 0,b,c+400,z+400 FROM corrupt_bit_test_ā;
      INSERT INTO corrupt_bit_test_ā SELECT 0,b,c+800,z+800 FROM corrupt_bit_test_ā;
      INSERT INTO corrupt_bit_test_ā SELECT 0,b,c+1600,z+1600 FROM corrupt_bit_test_ā;
      INSERT INTO corrupt_bit_test_ā SELECT 0,b,c+4000,z+4000 FROM corrupt_bit_test_ā;
      select count(*) from corrupt_bit_test_ā;
      count(*)
      1024
      CREATE INDEX idx3 ON corrupt_bit_test_ā(b, c);
      INSERT INTO corrupt_bit_test_ā VALUES(13000,'x',1,1);
       
      More results from queries before failure can be found in /var/lib/buildbot/maria-slave/power8-vlp04-debug/build/mysql-test/var/4/log/innodb_corrupt_bit.log

      I find it highly suspicious that both times the failure happened on P8. It's a plain test, there should be no race conditions involved, so there must be some other reason (build problems? environmental?). In any case, it's worth investigating.

        Attachments

          Issue Links

            Activity

              People

              • Assignee:
                jplindst Jan Lindström
                Reporter:
                elenst Elena Stepanova
              • Votes:
                0 Vote for this issue
                Watchers:
                3 Start watching this issue

                Dates

                • Created:
                  Updated:
                  Resolved: