Previously, the rejoin would only be ran on servers with a connected slave io
thread. This patch runs the rejoin also on slaves which cannot connect to a
downed old master while the master hostname or port differs from the current
cluster master server.
The service for a dummy session will be NULL. If authentication fails for
a dummy session, then no service level actions should be taken.
Only the binlogrouter can trigger authentication failure with a dummy
session as it creates connections before the service itself has started.
The fact whether a query explicitly defined a query was ignored by the
statement parsing function and it always assumed the database was
explicitly defined. This caused the TABLE keyword in a CREATE TABLE
statement to be falsely identified as the current database.
The readqueue should never be explicitly assigned and should only ever be
appended to. This guarantees that the packets are read and processed in
the correct order.
Also removed an unused function that deals with the readqueue
manipulation.
When packets were routed individually, the qualification for the
persistent pool was done before the current command was updated. In
addition to this, the previous commit doesn't seem like it can even build.
The :memory: database was misspelled as :memory without the trailing
colon. This caused an actual on-disk database to be created instead of an
in-memory one.
The thread-local user cache removes most of the cross-thread communication
from the user authentication at the cost of increased memory use and extra
network usage when users are loaded.
The statement tracking was given the same buffer multiple times when a
large packet was spread across multiple buffers when a router with
RCAP_TYPE_STMT_INPUT was present.
The command tracking packet processing is redundant for routers that
require RCAP_TYPE_STMT as the same processing is done later when the
buffer is split into packets and routed. By using the split packets, the
protocol module can simplify the command tracking by a great deal for most
routers.
When enabled, the monitor will redirect servers to replicate from the
current master. Standalone servers and servers replicating from a slave
are redirected.
MXS-1545: handling of slave file EOF refactoring.
Some slave/router state are now checked before any WARN/ERROR messages
about slave file EOF.
The missing “next_file” is always logged with warn.
When backend authentication failed due to errors other than wrong
credentials, the users were unconditionally reloaded. This caused a spike
of activity whenever authentication failed for other reasons.
Also fixed the test that checks for this to look for the correct error
message.
MXS-1530: check ANNOTATE_ROWS flag in connecting slave.
In MariaDB 10.2.4 replicate_annotate_row_events and
binlog_annotate_row_events have default to ON: this change checks
whether the connecting slave is not has ANNOTATE_ROWS in
blr_slave_binlog_dump(), those ANNOTATE_ROWS events can be sent or not
to the slave.
The new parameter allows ignoring of master servers that are external to
the monitor configuration. This allows sub-trees of the actual replication
tree to be used as fully fledged replication trees.
When a table map event is read after an alter table, the old TABLE_MAP
object contains old information. Due to this, as well as the added benefit
of making the code easier to read, the recycling of TABLE_MAP objects was
removed. In practice, there were no benefits to re-mapping the tables to a
different ID.
The Annotate_rows events were not processed which caused the following
table map event to be ignored.
Also removed a false debug assertion. The byte count can be zero and the
pointer is not guaranteed to point to anything valid.
If a CREATE TABLE statement had a quoted keyword as the name of a field,
the calculated column count and actual column counts would differ.
In addition to this, oneline comments before the end of the statement
would truncate the SQL due to the fact that the whitespace was squashed
before the comment removal was done.
Stop replicating from master if unsupported binlog events are seen.
Also report error message for unsupported events
(blr_read_events_all_events) at maxscale start-up and with
maxbinlogcheck utility
If the gtid_domain_pos of the master is ever modified,
gtid-variables will have multiple domains. Generally, we are
only interested in the most recent domain. This is tracked in
gtid_domain_id:s and the value of the master is used for
filtering the correct domain from all gtid-values.
Also, use gtid_current_pos instead of gtid_slave_pos. The
advantage of current_pos is that the same variable works also
for master servers. The gtid-handling is now more thorough and
detects some weird situations.
