This regression is caused by commit c380e16597 (MDEV-38833 "Suboptimal
or dead code at server startup"). Problem is that there should never
be an expectation that InnoDB bootstrap is crash-safe. Therefore,
after wsrep has initialized wsrep schema force checkpoint. This
ensured that InnoDB has successfully bootstrapped and wsrep chema
is initialized. This checkpoint ensures that wsrep-recover
is successful.
Other changes:
* Splitted galera_sst_mysqldump to two parts wsrep_sst_mysqldump
and wsrep_sst_mysqldump_debug that is run only on debug builds.
* Changed wserp_schema initialization not to use std::string
because it is unnecessary.
* Force checkpoint after wsrep_schema has been created.
* Make test wsrep_sst_mysqldump_with_key smaller
* Stabilize test cases
In the get_table_category function, the check for whether
a table is a WSREP system table was incorrectly placed.
As a result, TABLE_CATEGORY_MYSQL was returned, but user
access to tables in this category is not allowed when
the server is detached.
Fixed by moving the check for whether a table is a WSREP
system table inside the section with checks for MySQL
schema tables. Its category is now correctly set to
TABLE_CATEGORY_INFORMATION, and tables in this category
are accessible when detached.
Signed-off-by: Julius Goryavsky <julius.goryavsky@mariadb.com>
Problem was caused by MDEV-31413 commit 277968aa where
mysql.gtid_slave_pos table was replicated by Galera.
However, as not all nodes in Galera cluster are replica
nodes, rows were not deleted from table.
In this fix this is corrected so that mysql.gtid_slave_pos
table is not replicated by Galera. Instead when Galera
node receives GTID event and wsrep_gtid_mode=1, this event
is stored to mysql.gtid_slave_pos table.
Added test case galera_2primary_replica for 2 async
primaries replicating to galera cluster.
Added test case galera_circular_replication where
async primary replicates to galera cluster and
one of the galera cluster nodes is master
to async replica.
Modified test case galera_restart_replica to monitor
gtid positions and rows in mysql.gtid_pos_table.
Problem was that wsrep_schema tables were not marked as
category information. Fix allows access to wsrep_schema
tables even when node is detached.
This is 10.4-10.9 version of fix.
Signed-off-by: Julius Goryavsky <julius.goryavsky@mariadb.com>
Set mysql.wsrep_cluster and mysql.wsrep_cluster_members as
TABLE_CATEGORY_INFORMATION as mysql.wsrep_streaming_log
so that they can be queried even if node is not primary
component.
Signed-off-by: Julius Goryavsky <julius.goryavsky@mariadb.com>
* Removed all references related to wsrep_thd_pool (which was removed)
* Removed unused declarations in wsrep_schema.h
* The following would result invalid reads in
Wsrep_schema::replay_transaction():
```
frag_table->field[4]->val_str(&buf);
Wsrep_schema_impl::end_index_scan(frag_table);
Wsrep_schema_impl::finish_stmt(thd);
ret= wsrep_apply_events(thd, rli, buf.c_ptr_safe(), buf.length());
```
because `buf` was accessed after closing the table. The fix is to
perform storage reads using a different THD.
* In Wsrep_schema::recover_sr_transactions(), cluster_table was opened
for write, however it is only read here. And frag_table was opened
for read, wereas write is potentially needed.
Also, avoid copy caused by String::c_ptr() to zero terminate the c
string, use c_ptr_quick instead.
