| Bug #79331 | SYS_VAR.INNODB_MONITOR and INNODB_ZIP.16K,4K,8K TESTS ARE FAILING ON TRUNK | ||
|---|---|---|---|
| Submitted: | 18 Nov 2015 14:48 | Modified: | 20 Nov 2015 13:55 |
| Reporter: | Naga Satyanarayana Bodapati | Email Updates: | |
| Status: | Closed | Impact on me: | |
| Category: | MySQL Server: InnoDB storage engine | Severity: | S3 (Non-critical) |
| Version: | OS: | Any | |
| Assigned to: | CPU Architecture: | Any | |
[18 Nov 2015 15:20]
Naga Satyanarayana Bodapati
Posted by developer:
git log -1
commit bb8a1edcd77e8e8be2c9880af864972d84b26a68
Author: Satya Bodapati <satya.bodapati@oracle.com>
Date: Wed Nov 18 20:47:17 2015 +0530
Bug#22232974 - SYS_VAR.INNODB_MONITOR_* AND INNODB_ZIP.16K,4K,8K TESTS ARE FAILING ON TRUNK
Problem:
--------
SYS_VAR.INNODB_MONITOR_*.tests fail on trunk with the same reason mentioned by
this commit: 9519e3e98ed95e6301ad5fb7cf63746caaaef65e
" This test would fail when run as the first test after server startup,
because we seem to open the DD tables 'on demand' when executing the
first CREATE TABLE statement. If there were any previous tests executed
against the server instance, the DD tables would already have been
opened. "
INNODB_ZIP.16K,4K,8K tests fail because of this column removal from
mysql.tables by commit: e0f030fc638ab2b18cae13adf77d8c4729703765
Fix:
----
For innodb_monitor_*.test, use the same workaround. i.e. create and drop
a dummy table, so that the DD tables will not be opened during the
execution of the CREATE TABLE statement.
For innodb_zip.16k,8k,4k, just re-record the result.
Reviewed-By: Marko Makela <marko.makela@oracle.com>
Approved via IM (patch on bug page)
[20 Nov 2015 13:55]
Daniel Price
Posted by developer: This bug is test case related. No changelog entry required.

Description: sys_vars.innodb_monitor_reset_all_basic w4 [ fail ] Test ended at 2015-11-14 19:30:55 CURRENT_TEST: sys_vars.innodb_monitor_reset_all_basic --- /export/home2/pb2/test/sb_2-17062960-1447523288.36/usr/share/mysql-test/suite/sys_vars/r/innodb_monitor_reset_all_basic.result 2015-11-14 20:51:37.133390713 +0300 +++ /dev/shm/mtr-23085/var-normal-debug/4/log/innodb_monitor_reset_all_basic.reject 2015-11-14 21:30:55.167892683 +0300 @@ -327,14 +327,14 @@ from information_schema.innodb_metrics where name = "metadata_table_handles_opened"; name max_count min_count count max_count_reset min_count_reset count_reset status -metadata_table_handles_opened 12 NULL 12 12 NULL 12 enabled +metadata_table_handles_opened 1 NULL 1 1 NULL 1 enabled =============== innodb_zip.16k w2 [ fail ] Test ended at 2015-11-15 10:43:44 CURRENT_TEST: innodb_zip.16k --- /export/home/pb2/test/sb_2-17062512-1447573398.71/usr/share/mysql-test/suite/innodb_zip/r/16k.result 2015-11-15 10:46:24.584296054 +0300 +++ /dev/shm/mtr-27437/var-embedded-big/2/log/16k.reject 2015-11-15 12:43:43.364717940 +0300 @@ -90,12 +90,12 @@ mysql/slave_master_info 28 33 PRIMARY 3 3 1 50 mysql/slave_relay_log_info 12 33 PRIMARY 3 3 1 50 mysql/slave_worker_info 16 33 PRIMARY 3 3 2 50 -mysql/tables 34 33 PRIMARY 3 3 1 50 -mysql/tables 34 33 schema_id 4 2 2 50 -mysql/tables 34 33 engine 5 2 2 50 -mysql/tables 34 33 engine_2 6 0 1 50 -mysql/tables 34 33 collation_id 7 0 1 50 -mysql/tables 34 33 tablespace_id 8 0 1 50 +mysql/tables 33 33 PRIMARY 3 3 1 50 +mysql/tables 33 33 schema_id 4 2 2 50 +mysql/tables 33 33 engine 5 2 2 50 +mysql/tables 33 33 engine_2 6 0 1 50 +mysql/tables 33 33 collation_id 7 0 1 50 +mysql/tables 33 33 tablespace_id 8 0 1 50 mysql/tablespaces 9 33 PRIMARY 3 3 1 50 mysql/tablespaces 9 33 name 4 2 1 50 mysql/tablespace_files 7 33 tablespace_id 3 3 2 50 8k and 4k also fail with same mismatch How to repeat: Run these tests with repeat=10 and with --big-test Suggested fix: 1) Monitor test should get the same treatment from this commit: commit 9519e3e98ed95e6301ad5fb7cf63746caaaef65e Author: Marko Mäkelä <marko.makela@oracle.com> Date: Wed Sep 23 13:39:18 2015 +0300 WL#6378: Fix the innodb.monitor test. This test would fail when run as the first test after server startup, because we seem to open the DD tables 'on demand' when executing the first CREATE TABLE statement. If there were any previous tests executed against the server instance, the DD tables would already have been opened. Workaround: Before resetting the InnoDB counters, create and drop a dummy table, so that the DD tables will not be opened during the execution of the first CREATE TABLE statement that we are monitoring. Reviewed-by: Dmitry Lenev <dmitry.lenev@oracle.com> 2) innodb_zip.16k,8k,4k tests are failing because of this removal version column from mysql.tables table. So the number of columns count reduced from 34 to 33. Just re-record these tests.