MDEV-38073: Always use tx_read_only=false for Wsrep system threads

[mariadb-DenisProtivensky]

Unify setting variables for Wsrep schema threads:

• use common variable initialization for all Wsrep schema threads.
• simplify the code for replay & recovery of SR transactions.

Apparently, there was a separate issue with applier thread variables:
wsrep_plugins_post_init() would overwrite thd->variables for every
applier thread and forget to restore proper read only context.
Then, upon every server transaction termination,
trans_reset_one_shot_statistics() would set thread's read only context
to the one stored in thd->variables, thus spoiling the read only
context value for appliers.

[janlindstrom]

In my opinion setting on sql_plugin.cc looks incorrect.

[janlindstrom]

Can you explain why this needs to be here, assuming that tx_read_only=true is configuration doest this just overwrite it ?

[mariadb-DenisProtivensky]

This is a nasty one. This is the comment from here (MariaDB#4422):

Apparently, there was a separate issue with applier thread variables: wsrep_plugins_post_init() would overwrite thd->variables for every applier thread and forget to restore proper default isolation level. Then, upon every server transaction termination,
trans_reset_one_shot_statistics() would set thread's isolation level to the one stored in thd->variables, thus spoiling the isolation level value for appliers.

I didn't want to duplicate it, but I can add it to the commit description here as well. Hopefully, I found only two variables tx_read_only and tx_isolation with such a behavior.

[janlindstrom]

For wsrep, I would save original values and then set these and restore global default after use.

[mariadb-DenisProtivensky]

It's not needed. It's a thread created and belonging to the applier. We set all the variables like that.

[janlindstrom]

Did this disapear or it it set somewhere?

[janlindstrom]

These are like this as they are RAII scoped objects restoring original context. If they are really not needed, they should be removed from code or in my opinion there could be just one RAII scoped object.

[mariadb-DenisProtivensky]

The function is called only from the method of wsrep schema of the same name. In that method, a new THD object is created, used in the function and destroyed. This THD object is now initialized with wsrep_init_thd_variables() below, so that it has all the needed properties as other schema threads.

All these RAII objects did mostly the same, but they didn't account for future modifications like setting proper isolation (I will do it in another PR) or setting tx_read_only.

To sum up: I de-obfuscated this code and made all interactions explicit.

[mariadb-DenisProtivensky]

Also, we cannot remove these RAII classes as they are used by storage service THD calls, which for some reason doesn't have all the values set. But I agree that mostly all these RAII classes are used in conjunction and all together at a time. We might merge them one day if we figure out how to simplify other code parts.