Avoid relatively expensive THD::store_globals() for every query in the
threadpool. Use a lighter version instead, that only resets some thread
local storage variables(THD, mysys, PSI), avoids some calculationms
and caches syscall gettid (Linux only) in a thread_local variable.
Also simplify Worker_context use, with RAII.
Do not repeat yourself.
Instead of having the same DBUG_EXECUTE_IF code in threadpool and
thread-per-connection, add this code to setup_connection_thread_globals()
which is executed in all scheduling modes.
Prior to this patch, it is possible to access freed memory
(THD::event_scheduler) from tp_post_kill_notification().
With this patch, memory is freed only when THD is no more accessible
from other threads, i.e after it is removed from the thread_list.
Amend check for unread client data in threadpool.
THD::NET will have unread data, in case client uses compression, and
wraps multiple commands into a single compression packet
MariaDB C/C sends COM_STMT_RESET+COM_STMT_EXECUTE, and wraps it into
a single compressed packet, when compression is on, thus trying to use
compression and prepared statements against a threadpool-enabled server
will result into a hang, before this patch.
Due to restricted size of the threadpool, execution of client queries can
be delayed (queued) for a while. This delay was interpreted as client
inactivity, and connection is closed, if client idle time + queue time
exceeds wait_timeout.
But users did not expect queue time to be included into wait_timeout.
This patch changes the behavior. We don't close connection anymore,
if there is some unread data present on connection,
even if wait_timeout is exceeded. Unread data means that client
was not idle, it sent a query, which we did not have time to process yet.
This patch reduces the overhead of system calls prior to a query, for
threadpool. Previously, 3 system calls were done
1. WSARecv() to get notification of input data from client, asynchronous
equivalent of select() in one-thread-per-connection
2. recv(4 bytes) - reading packet header length
3. recv(packet payload)
Now there will be usually, just WSARecv(), which pre-reads user data into
a buffer, so we spared 2 syscalls
Profiler shows the most expensive call WSARecv(16%CPU) becomes 4% CPU,
after the patch, benchmark results (network heavy ones like point-select)
improve by ~20%
The buffer management was rather carefully done to keep
buffers together, as Windows would keeps the pages pinned
in memory for the duration of async calls.
At most 1MB memory is used for the buffers, and overhead per-connection is
only 256 bytes, which should cover most of the uses.
SSL does not yet use the optmization, so far it does not properly use
VIO for reads and writes. Neither one-thread-per-connection would get any
benefit, but that should be fine, it is not even default on Windows.
All changes (except one) is of type
thd->transaction. -> thd->transaction->
thd->transaction points by default to 'thd->default_transaction'
This allows us to 'easily' have multiple active transactions for a
THD object, like when reading data from the mysql.proc table
read TLS with my_thread_var
write TLS with set_mysys_var()
my_thread_var is no longer __attribute__ ((const)): this attribute
is simply incorrect here. Read gcc manual for more information.
sql/threadpool_generic.cc fails with that attribute.
MariaDB 10.4 was crashing when thread-handling was set to
pool-of-threads and wsrep was enabled.
There were two apparent reasons for the crash:
- Connection handling in threadpool_common.cc was missing calls to
control wsrep client state.
- Thread specific storage which contains thread variables (THR_KEY_mysys)
was not handled appropriately by wsrep patch when pool-of-threads
was configured.
This patch addresses the above issues in the following way:
- Wsrep client state open/close was moved in thd_prepare_connection() and
end_connection() to have common handling for one-thread-per-connection
and pool-of-threads.
- Thread local storage handling in wsrep patch was reworked by introducing
set of wsrep_xxx_threadvars() calls which replace calls to
THD store_globals()/reset_globals() and deal with thread handling
specifics internally.
Wsrep-lib was updated to version which relaxes internal concurrency
related sanity checks.
Rollback code from wsrep_rollback_process() was extracted to separate calls
for better readability.
Post rollback thread was removed as it was completely unused.
Plugin fixed to not lock the LOCK_operations when not active.
Server fixed to lock the LOCK_plugin less - do it once per
thread and then only if a plugin was installed/uninstalled.
stalls etc better.
- thread_pool_exact_stats - uses high precision timestamp for
the time when connection was added to the queue. This timestamp helps
calculating queuing time shown in I_S.THREADPOOL_QUEUES entries.
- If thread_pool_dedicated_listener is on, then each group will have its
own dedicated listener, that does not convert to worker.
With this variable on, the queueing time in I_S.THREADPOOL_QUEUES , and
actual queue size in I_S.THREADPOOOL_GROUPS will be more exact, since
IO request are immediately dequeued from poll, without delay.
Part of MDEV-19313.
Removed LOCK_connection_count.
Removed duplicate denied_connections and connection_errors_max_connection
increment from create_new_thread(). Another increment done by
CONNECT::close_with_error().
Simplified away CONNECT::thread_count_incremented. Now close_with_error()
is always called with connection_count incremented.
Part of MDEV-19515 - Improve connect speed
Code properly integrated into callers instead.
do_handle_one_connection(): no need to reset thd and thd->thread_stack
as they're not updated by cache_thread().
Part of MDEV-19515 - Improve connect speed
It was meaningful only for one-thread-per-connection scheduler anyway,
so call init_new_connection_handler_thread() directly from
handle_one_connection().
It was somewhat harmful for no-threads scheduler, because it'd attempt to
detach main thread on AIX_3_2 and Siemens unix (and DEC OSF/1 3.2 too).
Also it does duplicate my_thread_init() call, which may produce a warning
under EXTRA_DEBUG_THREADS.
start_wsrep_THD() shouldn't have called it at all: it already detached
and called my_thread_init().
Part of MDEV-19515 - Improve connect speed
