MySQL/MariaDB bugs
Update: MySQL 8.0.11 GA was released at the same time as I was writing this blog and still no sign is of the feature requested above. Waiting two years for the next MySQL version to implement that is a little long, I would be completely ok with this feature being implemented in a minor version of MySQL 8.0 (or even 5.7).
This January 2018, I spent some time testing MySQL and MariaDB parallel replication during a Booking.com Hack-a-thon (yes, there are three links in the word Hack-a-thon and there could even be more). I shared some of my results at the MySQL Pre-FOSDEM day and I will share more results during my tutorial at Percona Live. During those tests, I met the following bugs in MariaDB 10.2:
- MDEV-15135: SHOW GLOBAL STATUS can deadlock Optimistic Parallel Replication (so does a SELECT in INFORMATION_SCHEMA)
- MDEV-15152: Optimistic Parallel Replication deadlocks with START SLAVE UNTIL
- MDEV-15608: MariaDB sometimes crashes on rollback of a transaction with BLOBs
- Bug#89247: Deadlock with MTS when slave_preserve_commit_order = ON
The descriptions above are self-explanatory and I will give more details in my tutorial. A small note about the 3rd one: I do not know if this is a bug only affecting MariaDB or if it is a bug in upstream (MariaDB 10.2 is back-porting InnoDB from MySQL 5.7). I guess the fix will tell us more...
Percona Live Santa Clara 2018
As written above, I am giving a tutorial on MySQL and MariaDB parallel replication at Percona Live Santa Clara 2018. This tutorial will be the concatenation of two previous parallel replication talks with some more details (if you attended one of the previous talks, you will definitely learn something new and you might also be surprised by new content even if you attended both previous talks). If you do not have a conference ticket yet, you can use the discount code SeeMeSpeakPL18 that should give you 10% back on the conference price.
In my tutorial, I will again cover conservative and optimistic parallel replication in MariaDB. Below, you can see, from left to right, the commit rate of single-threaded, conservative and optimistic replication (I already shared this graph many times):
Results for MySQL 8.0 are also interesting, here is a graph of the speedups of one of my tests from January 2018 that I shared at FOSDEM (you can expect more details in the tutorial):
I also re-ran tests on optimistic parallel replication with MariaDB 10.2, here is a preview of what I will show in the tutorial (this has never been shared before):
I am looking forward to meeting you at Percona Live.
In my tutorial, I will again cover conservative and optimistic parallel replication in MariaDB. Below, you can see, from left to right, the commit rate of single-threaded, conservative and optimistic replication (I already shared this graph many times):
Results for MySQL 8.0 are also interesting, here is a graph of the speedups of one of my tests from January 2018 that I shared at FOSDEM (you can expect more details in the tutorial):
I also re-ran tests on optimistic parallel replication with MariaDB 10.2, here is a preview of what I will show in the tutorial (this has never been shared before):
I am looking forward to meeting you at Percona Live.
No comments:
Post a Comment