Problem is that FLUSH TABLES WITH READ LOCK first blocks threads from
starting new commits, then waits for running commits to complete. But
in-order parallel replication needs commits to happen in a particular
order, so this can easily deadlock.
To fix this problem, this patch introduces a way to temporarily pause
the parallel replication worker threads. Before starting FTWRL, we let
all worker threads complete in-progress transactions, and then
wait. Then we proceed to take the global read lock. Once the lock is
obtained, we unpause the worker threads. Now commits are blocked from
starting by the global read lock, so the deadlock will no longer occur.
A few tests assumes that the CYCLE timer is always available,
which is not true on some platforms (e.g. ARM).
Fixing the tests not to reply on the CYCLE availability.
Moved Apc_target::destroy(), Apc_target::enable() and Apc_targe::disable()
definitions to my_apc.h so that they can be inlined.
Apc_targe::disable() now calls Apc_target::process_apc_requests() only if
there're APC requests. This saves one pthread_mutex_lock() call.
Overhead change:
Apc_target::disable 0.04% -> out of radar
Apc_target::enable 0.03% -> out of radar
Apc_target::process_apc_requests 0.02% -> out of radar
pthread_mutex_lock 0.43% -> 0.42%
pthread_mutex_unlock 0.26% -> 0.25%
table
Performance schema discovery fails if connection has no active database set.
This happened due to restriction in SQL parser: table name with no database name
is ambiguous in such case.
Fixed by temporary substitution of default database with being discovered table
database.
including the big commit
commit 305130361bf72726de220f3d2b2787395e10be61
Author: Marc Alff <marc.alff@oracle.com>
Date: Tue Feb 10 11:31:32 2015 +0100
WL#8354 BACKPORT DIGEST IMPROVEMENTS TO MYSQL 5.6
(with the following commits) and related changes in sql/
XID cache is now based on lock-free hash.
Also fixed lf_hash_destroy() to call alloc destructor.
Note that previous implementation had race condition when thread was accessing
XA owned by different thread. This new implementation doesn't fix it either.
During slow execution, e.g. under valgrind, there was a chance
that Aria checkpoint would happen while P_S tables were being
queried; it could cause different data in joined P_S, and
thus combinations of results that the test did not expect.
Fixed by disabling Aria checkpoints for the test.
Implement --semi-sync-master-wait-point=AFTER_SYNC|AFTER_COMMIT.
When AFTER_SYNC, the semi-sync wait will be done earlier, before the storage
engine commit rather than after. This means that a transaction will not be
visible on the master until at least one slave has received it.
Fix a possible race in the test case when restarting the server.
Make sure we have disconnected before waiting for the reconnect
that signals that the server is back up. Otherwise, we may in
rare cases continue the test while the old server is shutting
down, eventually leading to "connection lost" failure.
