InnoDB is the most similar storage engine for an Oracle developer. Understanding specifics such as the clustered primary key, secondary indexes sizes, page fill factor options and TEXT/BLOB management are important design needs developers need to know. Being able to monitor and analyze the available instrumentation and learn specifics for example auto-inc or gap locking contention, thread concurrency management options and data/log I/O options are also key in mastering InnoDB diagnostics.
4. the special sauce! clustered indexing natural order page directory www.BlueGecko.net
5. two short operational diversions innodb_flush_method = O_DIRECT is worth benchmarking on your system. choose smart primary keys www.BlueGecko.net
6. indexing curiosities secondary indexes point to the clustered index BLOB, VARCHAR, TEXT can be stored in overflow pages ERROR HY000: Too big row. www.BlueGecko.net
7. scaling issues when it’s time to go big,where are the bottlenecks? www.BlueGecko.net
14. operational concerns mysql> show innodbstatus; this gives you snapshot information trending gives you framework to see when things are going wrong http://code.google.com/p/mysql-cacti-templates www.BlueGecko.net
21. Blue Gecko and contact info sarah@bluegecko.net sarah.novotny@gmail.com @sarahnovotny @bluegecko senk on #mysql Blue Gecko provides Remote DBA services for companies around the world 7x24x365 support including monitoring, performance analysis, proactive maintenance and architectural guidance for small and large datasets. www.BlueGecko.net
Hinweis der Redaktion
table creation information in .frmMySQL’s core table locking is disabled (politely)transactions do not behave quite the same in innodb – autocommit on by default. transaction isolation == repeatable readdeadlock detection
b-tree primary key ++ record in index (not pointer)instead of moving ½ the records, data is added to the tablespaces and indexes in natural orderpage directory uses binary searching in sparse array fill factors are auto managed – and at my last check was still an open questionindex-organized table
since innodb buffers both key and data there are cases where O_DIRECT can help a lot. the idea being to avoid double cachingthere will be an arbitrary useless key that is assigned if you don’t set one!
secondary indexes point to clustered. no data. extra callSecondary index records must always fit on a single B-tree pageIn a clustered index, BLOB, VARCHAR and TEXT columns that are not part of the primary key may be stored on separately allocated (“overflow”) pages. We call these “off-page columns” whose values are stored on singly-linked lists of overflow pages. ERROR HY000: Too big row. InnoDB requires that each B-tree page can accommodate at least two records.
Much performance pain come into play with lock contentionglobal counter which held the last value for the auto-incrementing column. A lock would be placed on this counter for the duration of the SQL statement which did the inserting. until MySQL 5.1.22new server variable, innodb_autoinc_lock_mode -- limits how often global locking is used and still defaults to that if not able to preserve derterministicnessA type of gap lock called an insertion intention gap lock is set by INSERT operations prior to row insertion.next key locks combo of record lock and gap lock.
Much performance pain come into play with lock contentionNames changed to protect the innocent
can’t reprise or due justice to this talk.DomasMituzas – FB go watch it.oreilly’smysql conference 2010
not enough knobs, must benchmark.innodb_thread_concurrency
LVM and slaves give easy opportunities for creating clones for testinginnodbhotbackupxtra backup
You might see these graphs spike during times of high concurrency or contention. These graphs basically indicate different types of activity involved in obtaining row locks or mutexes, both of which are causes of poor scaling in standard InnoDBXtraDB, the InnoDBplugin, and the Percona-patched MySQL are improved.
different types of activity involved in obtaining row locks or mutexesadditionally, can see what’s going on in individual transactions.