The unsung heroes of InnoDB are the logfiles. They are what makes InnoDB automatic crash recovery possible.

Database administrators of other DBMS may be familiar with the concept of a “redo” log. When data is changed, affected data pages are changed in the innodb_buffer_pool. Then, the change is written to the redo log, which in MySQL is the InnoDB logfile (ib_logfile0 and ib_logfile1). The pages are marked as “dirty”, and eventually get flushed and written to disk.

If MySQL crashes, there may be data that is changed that has not been written to disk. Those data pages were marked as “dirty” in the innodb_buffer_pool, but after a MySQL crash the innodb_buffer_pool no longer exists. However, they were written to the redo log. On crash recovery, MySQL can read the redo log (InnoDB log files) and apply any changes that were not written to disk.

That is the basic functionality of the InnoDB log files. Given this, …

Ever wonder what is stored in InnoDB buffer pool at the moment ?

It is not so hard actually - we made a short patch for MySQL 5.0 which show innodb buffer pool content

PLAIN TEXT SQL:

1. mysql> SELECT * FROM information_schema.INNODB_BUFFER_POOL_CONTENT; 
2. +-----------+-------+--------+---------+----------+------------+-----------+--------------+--------------+----------------+-----------------+--------------+------------------+
3. | BLOCK_NUM | SPACE | OFFSET | RECORDS | DATASIZE | FLUSH_TYPE | FIX_COUNT | LRU_POSITION | PAGE_TYPE_ID | PAGE_TYPE      | INDEX_NAME      | TABLE_SCHEMA | TABLE_NAME       |
4. …

In the previous post we had a look at the MySQL Cluster NDB API and how it enables direct access to the MySQL Cluster data nodes, and therefore also enables access through other protocols than SQL.

I've often asked myself: Since NDB is so great for MySQL Cluster, is there anything similar for MySQL Server (the not-cluster version...)? A couple of months ago Kazuho Oku did something like that and wrote in his blog about it.

The context for Kazuho's work is the social network use case: 1) You have users. 2) Some users are linked to each others as friends. 3) When a user logs in, he should see a timeline of events/messages from his friends. In a previous post he had already tested the difference between a …

In the previous post we had a look at the MySQL Cluster NDB API and how it enables direct access to the MySQL Cluster data nodes, and therefore also enables access through other protocols than SQL.

I've often asked myself: Since NDB is so great for MySQL Cluster, is there anything similar for MySQL Server (the not-cluster version...)? A couple of months ago Kazuho Oku did something like that and wrote in his blog about it.

The context for Kazuho's work is the social network use case: 1) You have users. 2) Some users are linked to each others as friends. 3) When a user logs in, he should see a timeline of events/messages from his friends. In a previous post he had already tested the difference between a …

In the previous post we had a look at the MySQL Cluster NDB API and how it enables direct access to the MySQL Cluster data nodes, and therefore also enables access through other protocols than SQL.

I've often asked myself: Since NDB is so great for MySQL Cluster, is there anything similar for MySQL Server (the not-cluster version...)? A couple of months ago Kazuho Oku did something like that and wrote in his blog about it.

The context for Kazuho's work is the social network use case: 1) You have users. 2) Some users are linked to each others as friends. 3) When a user logs in, he should see a timeline of events/messages from his friends. In a previous post he had already tested the difference between a …

Before I leave Grazr to work at Lycos, I've been tasked with documenting several things-- MySQL setup, including replication, nagios, UDFs we use, etc. I was describing on our wiki what using InnoDB means, and some basic things about settings. Here's a tidbit of it that I think is useful information.

Some important InnoDB Characteristics

InnoDB stores data and files in the same place whereas MyISAM has separate index and data files. InnoDB stores it's data in either a single tablespace file (ibdataN -- n being number) which containins all tables, or a single auto-extending tablespace file set by innodb_file_per_table (tablename.ibd), which is what we use for grazr because file-per-table allows OPTIMIZE TABLE to reclaim space from deletions, would could gives better performance with better optimized tables. Also allows you to restore backups of single tables without interrupting the use of the remaining InnoDB tables (per …

To have online backups of MySQL, We recently bought a license for InnoBase/Oracle's InnoDB Hot Backup Tool, ibbackup. This tool, used in conjunction with innobackup, has worked great in creating a nightly backup, with no downtime during the backup. Not even nagios messages!

I run innobackup/ibbackup on one of our slaves (well, it's also a dual master, but not used by apps). innobackup produces a backup in a directory that I specified, and when run results in a time-stamped directory, as show below:


ls -l 2008-09-17_03-00-03/
total 276272
-rw-r--r-- 1 root root 349 2008-09-17 03:00 backup-my.cnf
drwxr-x--- 2 root root 4096 2008-09-17 03:55 grazr
-rw-r--r-- 1 root root 27 2008-09-17 03:55 ibbackup_binlog_info
-rw-r----- 1 root root 186109952 2008-09-17 03:55 ibbackup_logfile
-rw-r----- 1 root root 10485760 2008-09-17 03:00 ibdata1
-rw-r----- 1 root root …

I just got back from vacation and noticed that two patchsets have been released that greatly improve Innodb performance! Maybe I need to take more breaks

Yasufumi Kinoshita's patches Percona recently released a patch which includes performance fixes developed by Yasufumi Kinoshita from NTT Comware. This helps diskbound applications quite significantly. Details at Bug #29413 Maximum performance of OLTP benchmark is not so scalable on multi-cpu. It looks like the bulk of performance improvements come from breaking up the lock guarding the buffer pool structures, and improvements in the IO code path. The "buf_pool->mutex" also gets quite hot when concurrency is not limited (via innodb_thread_concurrency) and …

I just got back from vacation and noticed that two patchsets have been released that greatly improve Innodb performance! Maybe I need to take more breaks

Yasufumi Kinoshita's patches Percona recently released a patch which includes performance fixes developed by Yasufumi Kinoshita from NTT Comware. This helps diskbound applications quite significantly. Details at Bug #29413 Maximum performance of OLTP benchmark is not so scalable on multi-cpu. It looks like the bulk of performance improvements come from breaking up the lock guarding the buffer pool structures, and improvements in the IO code path. The "buf_pool->mutex" also gets quite hot when concurrency is not limited (via innodb_thread_concurrency) and …

I just got back from vacation and noticed that two patchsets have been released that greatly improve Innodb performance! Maybe I need to take more breaks

Yasufumi Kinoshita's patches Percona recently released a patch which includes performance fixes developed by Yasufumi Kinoshita from NTT Comware. This helps diskbound applications quite significantly. Details at Bug #29413 Maximum performance of OLTP benchmark is not so scalable on multi-cpu. It looks like the bulk of performance improvements come from breaking up the lock guarding the buffer pool structures, and improvements in the IO code path. The "buf_pool->mutex" also gets quite hot when concurrency is not limited (via innodb_thread_concurrency) and …