The documentation of innodb_fake_changes is not very clear

[Ovais Tariq]

Hi,

The documentation of innodb_fake_changes here http://www.percona.com/doc/percona-server/5.5/management/innodb_fake_changes.html is not very clear about its usage. From reading the docs it looks as though the slave transparently starts prefetching when innodb_fake_changes is enabled, while that is certainly not the case.

[Ovais Tariq (ovais-tariq)]

I also think that it should be more prominently mentioned that this feature is only safe to use in an InnoDB-only server, because MyISAM changes cannot be rolled back.

[Ovais Tariq (ovais-tariq)]

BTW the two fake changes related variables listed in the documentation http://www.percona.com/doc/percona-server/5.5/management/innodb_fake_changes.html have their scope listed as GLOBAL, however the scope is both GLOBAL as well as SESSION level.

[Ovais Tariq (ovais-tariq)]

I also think that the description about "DDL operations are not supported" should also be changed, as its quite confusing.

[Raghavendra D Prabhu (raghavendra-prabhu)]

The 'data may be lost' may need to be further elucidated.

I presume since DDL operations fail with HA_ERR_WRONG_COMMAND,
subsequent DML operations may fail (from missing column etc.),
hence termed as data lost.

If there is any other data loss possibility with any of the
caveats, it should be highlighted.

[Ovais Tariq (ovais-tariq)]

Any ideas when this will get fixed. Its almost 11 months :)

[Hrvoje Matijakovic (hrvojem)]

For 5.5 fixed here: https://github.com/percona/percona-server/pull/199

[Hrvoje Matijakovic (hrvojem)]

For 5.6 fixed here: https://github.com/percona/percona-server/pull/201

[Shahriyar Rzayev (rzayev-sehriyar)]

Percona now uses JIRA for bug reports so this bug report is migrated to: https://jira.percona.com/browse/PS-639