Friday, April 20, 2018

Fun with Bugs #67 - On Some Public Bugs Fixed in MySQL 8.0.11 GA

I stopped reviewing MySQL Release Notes for quite a some time, but major GA releases of MySQL do not happen often, so I decided to make an exception and write about some bugs from Community users fixed in MySQL 8.0.11 GA.

I'll start with good news about MySQL 8.0.11 GA! You can get sources at GitHub, and I had no problems to build on Fedora 27 on my good old QuadCore box, using the following cmake command line:
[openxs@fc23 mysql-server]$ cmake . -DCMAKE_BUILD_TYPE=RelWithDebInfo -DBUILD_CONFIG=mysql_release -DFEATURE_SET=community -DWITH_EMBEDDED_SERVER=OFF -DDOWNLOAD_BOOST=1 -DWITH_BOOST=/home/openxs/boost -DENABLE_DOWNLOADS=1 -DWITH_UNIT_TESTS=OFF -DCMAKE_INSTALL_PREFIX=/home/openxs/dbs/8.0
[openxs@fc23 mysql-server]$ time make -j 4

[100%] Built target mysqld

real    33m52.791s
user    105m47.475s
sys     8m19.018s
Comparing to previous experience, I had minor problem with unit tests, so just skipped them with -DWITH_UNIT_TESTS=OFF option. There is no problem to run the resulting binaries, unless you try to use data directory from older 8.0.x. Then you'll end up with:
2018-04-19T15:36:35.165841Z 1 [ERROR] [MY-011092] [Server] Upgrading the data dictionary from dictionary version '80004' is not supported.
2018-04-19T15:36:35.166239Z 0 [ERROR] [MY-010020] [Server] Data Dictionary initialization failed.
2018-04-19T15:36:35.166310Z 0 [ERROR] [MY-010119] [Server] Aborting
I had to remove data directory and initialize it from scratch (it was testing instance anyway, last time used for real while I worked on this presentation):
[openxs@fc23 8.0]$ rm -rf data/*
[openxs@fc23 8.0]$ bin/mysqld --no-defaults --initialize-insecure --port=3308 --socket=/tmp/mysql.sock --basedir=/home/openxs/dbs/8.0 --datadir=/home/openxs/dbs/8.0/data --skip-log-bin
2018-04-19T15:43:55.324606Z 0 [Warning] [MY-010139] [Server] Changed limits: max_open_files: 1024 (requested 8161)
2018-04-19T15:43:55.324726Z 0 [Warning] [MY-010142] [Server] Changed limits: table_open_cache: 431 (requested 4000)
2018-04-19T15:43:55.325147Z 0 [System] [MY-013169] [Server] /home/openxs/dbs/8.0/bin/mysqld (mysqld 8.0.11) initializing of server in progress as process 20034
2018-04-19T15:44:14.438776Z 4 [Warning] [MY-010453] [Server] root@localhost is created with an empty password ! Please consider switching off the --initialize-insecure option.
2018-04-19T15:44:29.625227Z 0 [System] [MY-013170] [Server] /home/openxs/dbs/8.0/bin/mysqld (mysqld 8.0.11) initializing of server has completed
[openxs@fc23 8.0]$ bin/mysqld_safe --no-defaults --port=3308 --socket=/tmp/mysql.sock --basedir=/home/openxs/dbs/8.0 --datadir=/home/openxs/dbs/8.0/data --skip-log-bin &
[1] 20080
[openxs@fc23 8.0]$ 2018-04-19T15:44:58.224816Z mysqld_safe Logging to '/home/openxs/dbs/8.0/data/fc23.err'.
2018-04-19T15:44:58.271255Z mysqld_safe Starting mysqld daemon with databases from /home/openxs/dbs/8.0/data

[openxs@fc23 8.0]$ bin/mysql -uroot --socket=/tmp/mysql.sock
Welcome to the MySQL monitor.  Commands end with ; or \g.
Your MySQL connection id is 7
Server version: 8.0.11 MySQL Community Server (GPL)

Copyright (c) 2000, 2018, Oracle and/or its affiliates. All rights reserved.

Oracle is a registered trademark of Oracle Corporation and/or its
affiliates. Other names may be trademarks of their respective

Type 'help;' or '\h' for help. Type '\c' to clear the current input statement.

mysql> show variables like '%version%';
| Variable_name           | Value                        |
| innodb_version          | 8.0.11                       |
| protocol_version        | 10                           |
| slave_type_conversions  |                              |
| tls_version             | TLSv1,TLSv1.1,TLSv1.2        |
| version                 | 8.0.11                       |
| version_comment         | MySQL Community Server (GPL) |

| version_compile_machine | x86_64                       |
| version_compile_os      | Linux                        |
| version_compile_zlib    | 1.2.11                       |
9 rows in set (0.00 sec)
So, you can build MySQL 8.0.11 right now and start using it to make your own conclusions about this release.

I still do not care about NoSQL, JSON, new cool features etc. You'll see megabytes of texts about these by the end of 2018. I am going to concentrate mostly on InnoDB, replication bugs and few others:
  •  I am happy to start with Bugt #89509 - "Valgrind error on innodb.blob_page_reserve, bundled zlib", reported by Laurynas Biveinis. See also his Bug #89597 - "Valgrind reporting memory leak on MTR test main.validate_password_component" and Bug #89433 - "NULL dereference in dd::tables::DD_properties:unchecked_get". Percona engineers spent a lot of efforts recently testing MySQL 8.0.x and reporting bugs noted. I think Oracle should explicitly admit the impact of Percona's QA effrots for the quality of this GA release.
  • Biug #89127 - "Optimize trx_rw_is_active() by tracking the lowest active transaction id". This bug was reported by Zhai Weixiang, who had suggested a patch also.
  • Bug #89129 - "create table+DML on innodb_ddl_log table=crash in". This bug was reported by Ramana Yeruva. Tables were made protected and DDL and DML operations on these tables are no longer permitted.
  • Bug #89087 - "Assertion `key->flags & 1' failed". This debug assertion (related to the way PRIMARY key was created based on UNIQUE one) was reported by Roel Van de Paar for 5.7.21, but we see the fix documented only for 8.0.x.
  • Bug #87827 - "Performance regression in "create table" speed and scalability in 8.0.3". It was reported by Alexander Rubin from Percona.
  • Bug #87812 - "Concurrent DDL operation in progress even after acquiring backup lock". Nice bug report from Debarun Banerjee.
  • Bug #87532 - "Replay log record cause mysqld crash during online DDL". I am happy to see impovements in "online ALTER" implementation that covers all GA versions, not just 8.0. I am also happy to see Oracle engineers (Ohm Hong in this case) reporting bugs in public!
  • Bug #88272 - "Assertion `new_value >= 0' failed.". Yet another debug assertion found by Roel Van de Paar, this time related to GTIDs and XA transactions. Check also his Bug #88262 - "ERROR 1598 (HY000): Binary logging not possible + abort".
  • Bug #84415 - "slave don't report Seconds_Behind_Master when running slave_parallel_workers > 0". Yet another contribution from Percona engineers. This bug was reported by Marcelo Altmann and patches were provided by Robert Golebiowski. This bug is also fixed in MySQL 5.7.22.
  • Bug #89793 - "INFORMATION_SCHEMA.STATISTICS field type change". Unexpected change in early 8.0.x versions was noted and reported by Mark Guinness.
  • Bug #89584 - "5.7->8.0 upgrade crash with default-time-zone set". Nice to see this bug (reported by Shane Bester) fixed in GA release.
  • Bug #89487 - "ALTER TABLE hangs in "Waiting for tablespace metadata lock" state". This regression bug was reported by Sveta Smirnova.
  • Bug #89324 - "main.comment_column2 fails with compression". This regression was noted and reported by Manuel Ung.
  • Bug #89122 - "Severe performance regression in server bootstrap". I am really happy to see this bug reported by Georgi Kodinov fixed. I noted it as soon as I started testing 8.0.x (see a duplicate by Roel Van de Paar, Bug #89444) and it was very annoying. I've already checked (see above) that the problem is gone!
  • Bug #89038 - "Add new column to 'mysql.routines' to accommodate the Polygot project". So, Oracle is planning to support stored programs in different languages! Thank you, Sivert Sørumgård, for reporting this in public! See also his Bug #89035 - "Reject LCTN changing after --initialize".
  • Bug #87836 - "XA COMMIT/ROLLBACK rejected by non-autocommit session with no active transaction". It would be sad if this bug is not fixed in MySQL 5.7.x, where it was originally found by Wei Zhao.
  • Bug #87708 - "MDL for column statistics is not properly reflected in P_S.METADATA_LOCKS". It was reported by Erik Frøseth.
  • Bug #85997 - "inplace alter table with foreign keys causes table definition mismatch". This bug was reported by Magnus Blåudd.
  • Bug #85561 - "Users can be assigned non-existing roles as default". Nice to see this bug reported by Giuseppe Maxia fixed in GA release.
  • Bug #33004 - "integer constants casted to bigints by unions". This bug was reported by Domas Mituzas more than 10 years ago!
Now I have to stop, as I found private bug in release notes, Bug #89512. Based on description:
"Window function row-buffer handling has been refactored to reduce the number of handler reads by 25%. (Bug #89512, Bug #27484133)"
I truly do not get why it remains private (or why it was reported in public for such a "sensitive" matter), so I better stop.

MySQL 8 is GA, finally! There are a lot more fixes there that I had not mentioned above. I am surely there is even more bugs to find. So, happy hunting!

1 comment:

  1. 8.0.11 looks awesome but caching_sha2_password is going to cause some confusion -