Rename XtraDB options

[Vadim Tkachenko]

Next XtraDB options may be renamed in Percona Server 5.5

innodb_ibuf_active_contract
innodb_ibuf_accel_rate
innodb_pass_corrupt_table
innodb_expand_import

[Vadim Tkachenko (vadim-tk)]

Assigned to Baron to provide new names

[Baron Schwartz (baron-xaprb)]

 innodb_adaptive_checkpoint
   This variable isn't named well. The InnoDB option innodb_adaptive_flushing
   is the correct name. Checkpointing isn't what this controls -- flushing is
   what it controls. We should simply support new values for
   innodb_adaptive_flushing instead of adding a new option.
 innodb_expand_import
   This variable is probably better named something else, but I think it's not
   so bad that we have to break compatibility for it. If we want to rename it,
   we could name it innodb_import_foreign_tablespaces.
 innodb_flush_log_at_trx_commit_session
   This variable shouldn't exist. Can we make the existing variable
   have both session and global behavior, like most variables?
 innodb_ibuf_accel_rate
 innodb_ibuf_active_contract
 innodb_ibuf_max_size
   We should rename these variables to match the InnoDB name,
   innodb_change_buffering. I suggest:
    innodb_ibuf_accel_rate = innodb_change_buffer_purge_rate
    innodb_ibuf_active_contract = innodb_change_buffer_purge_when = <idle,always>
    innodb_ibuf_max_size = innodb_change_buffer_max_size
 innodb_pass_corrupt_table
   This should be innodb_corruption_detection = <warn,crash>.
 innodb_use_purge_thread
   This should be innodb_purge_threads = <number>
 profiling_server
   This should a new value for log_slow_verbosity instead of a new option.
 suppress_log_warning_1592
   This should be called log_warnings_silence = <list,of,error,numbers>
 use_global_log_slow_control
 use_global_long_query_time
   I would suggest that we try to deprecate and remove these, and implement
   https://bugs.launchpad.net/percona-server/+bug/588854 instead.

In general, I suggest that in the future we avoid adding 0|1 options, and add options that take a string or list of strings. This will make it much easier to avoid renames in the future. For example, innodb_flush_logs_at_trx_commit was originally a boolean, but then they discovered they wanted a new value, so we have 0, 1, and 2. If it had been a string variable from the start, it could have been <background|always> and then they could have added a new value that was more descriptive than "2". We will find the same problem with yes/no options that we add, in my opinion.

[Yasufumi Kinoshita (yasufumi-kinoshita)]

comments for innodb_xxxx for now.

----
I think XtraDB should keep 100% compatibility to the exist names of the options.
It means don't change them.

Changing "innodb_adaptive_flushing" is not acceptable for me.
Should be added like "innnodb_adaptive_flushing_method" new value instead of "innodb_adaptive_checkpoint".

innnodb_adaptive_flushing_method = [native|estimate|keep_average]

----
innodb_flush_log_at_trx_commit also.

At https://bugs.launchpad.net/percona-server/+bug/635399
I suggested
innodb_use_global_flush_log_at_trx_commit = [0|1]
session value. This is acceptable to implement

----
I think the word "purge" is not be used for insert(change now?)_buffer. it is different from about rseg. Avoiding confusing.
in the innoDB source the word is "contract".
Do you know "purge" is used as officially in the innoDB manual/textbook ?

[Shahriyar Rzayev (rzayev-sehriyar)]

Percona now uses JIRA for bug reports so this bug report is migrated to: https://jira.percona.com/browse/PS-1176