Speaking at Oracle OpenWorld 2014

For those of you lucky enough to come and listen to all the great talks within the MySQL Central @ Oracle OpenWorld this year, I’ll also be giving a talk about the MySQL sys schema this year.

This builds upon the talks that I’ve given in the past around Performance Schema, as well as some of the great looking talks on Performance Schema that are on the schedule this year by other MySQL @ Oracle engineers (see below).

The schedule builder is now live, here are some of the other interesting ones on my list that I’d like to attend:

And of course, if you’re in to monitoring as much as me (and why wouldn’t you be?), be sure to check these out:

Hope to see you there – come find me, and lets have a beverage of your choice together!

MySQL sys version 1.1.0 released

I’ve just released the 1.1.0 version of the MySQL sys schema.

This release is hugely pleasing to me, in that I actually didn’t have to do too much work on it myself! There were a significant number of contributions from Jesper Wisborg Krogh and Arnaud Adant, both MySQL Support Engineers (at the time at least, Arnaud has moved on to pastures new now), as well as again from Joe Grasse.

Thank you all for your contributions!

Here’s a summary of the changes:

Improvements

  • Added host summary views, which have the same structure as the user summary views, but aggregated by host instead (Contributed by Arnaud Adant)
    • host_summary
    • host_summary_by_file_io_type
    • host_summary_by_file_io
    • host_summary_by_statement_type
    • host_summary_by_statement_latency
    • host_summary_by_stages
    • waits_by_host_by_latency
  • Added functions which return instruments are either enabled, or timed by default (#15) (Contributed by Jesper Wisborg Krogh)
    • ps_is_instrument_default_enabled
    • ps_is_instrument_default_timed
  • Added a ps_thread_id function, which returns the thread_id value exposed within performance_schema for the current connection (Contributed by Jesper Wisborg Krogh)
  • Improved each of the user specific views to return aggregate values for background threads, instead of ignoring them (Contributed by Joe Grasse)
  • Optimized the schema_table_statistics and schema_table_statistics_with_buffer views, to use a new view that will get materialized (x$ps_schema_table_statistics_io), along with the changes to the RETURN types for extract_schema_from_file_name and extract_table_from_file_name, this results in a significant performance improvement – in one test changing the run time from 14 minutes to 20 seconds. (Conceived by Roy Lyseng, Mark Leith and Jesper Wisborg Krogh, implemented and contributed by Jesper Wisborg Krogh)

Bug Fixes

  • Removed unintentially committed sys_56_rds.sql file (See Issue #5, which is still outstanding)
  • Fixed the ps_trace_statement_digest and ps_trace_thread procedures to properly set sql_log_bin, and reset the thread INSTRUMENTED value correctly (Contributed by Jesper Wisborg Krogh)
  • Removed various sql_log_bin disabling from other procedures that no longer require it – DML against the performance_schema data is no longer replicated (Contributed by Jesper Wisborg Krogh)
  • Fixed EXPLAIN within ps_trace_statement_digest procedure (Contributed by Jesper Wisborg Krogh)
  • Fixed the datatype for the thd_id variable within the ps_thread_stack procedure (Contributed by Jesper Wisborg Krogh)
  • Fixed datatypes used for temporary tables within the ps_trace_statement_digest procedure (Contributed by Jesper Wisborg Krogh)
  • Fixed the RETURN datatype extract_schema_from_file_name and extract_table_from_file_name to return a VARCHAR(64) (Contributed by Jesper Wisborg Krogh)
  • Added events_transactions_current to the default enabled consumers in 5.7 (#25)